Enhanced mobility management mechanisms for 5G networks

Enhanced mobility management mechanisms for 5G Networks Akshay Jain ADVERTIMENT La consulta d’aquesta tesi queda condicionada a l’acceptació de les següents condicions d'ús: La difusió d’aquesta tesi per mitjà del r e p o s i t o r i i n s t i t u c i o n a l UPCommons (http://upcommons.upc.edu/tesis) i el repositori cooperatiu TDX ( h t t p : / / w w w . t d x . c a t / ) ha estat autoritzada pels titulars dels drets de propietat intel·lectual únicament per a usos privats emmarcats en activitats d’investigació i docència. No s’autoritza la seva reproducció amb finalitats de lucre ni la seva difusió i posada a disposició des d’un lloc aliè al servei UPCommons o TDX. No s’autoritza la presentació del seu contingut en una finestra o marc aliè a UPCommons (framing). Aquesta reserva de drets afecta tant al resum de presentació de la tesi com als seus continguts. En la utilització o cita de parts de la tesi és obligat indicar el nom de la persona autora. ADVERTENCIA La consulta de esta tesis queda condicionada a la aceptación de las siguientes condiciones de uso: La difusión de esta tesis por medio del repositorio institucional UPCommons (http://upcommons.upc.edu/tesis) y el repositorio cooperativo TDR (http://www.tdx.cat/?localeattribute=es) ha sido autorizada por los titulares de los derechos de propiedad intelectual únicamente para usos privados enmarcados en actividades de investigación y docencia. No se autoriza su reproducción con finalidades de lucro ni su difusión y puesta a disposición desde un sitio ajeno al servicio UPCommons No se autoriza la presentación de su contenido en una ventana o marco ajeno a UPCommons (framing). Esta reserva de derechos afecta tanto al resumen de presentación de la tesis como a sus contenidos. En la utilización o cita de partes de la tesis es obligado indicar el nombre de la persona autora. WARNING On having consulted this thesis you’re accepting the following use conditions: Spreading this thesis by the institutional repository UPCommons (http://upcommons.upc.edu/tesis) and the cooperative repository TDX (http://www.tdx.cat/?localeattribute=en) has been authorized by the titular of the intellectual property rights only for private uses placed in investigation and teaching activities. Reproduction with lucrative aims is not authorized neither its spreading nor availability from a site foreign to the UPCommons service. Introducing its content in a window or frame foreign to the UPCommons service is not authorized (framing). These rights affect to the presentation summary of the thesis as well as to its contents. In the using or citation of parts of the thesis it’s obliged to indicate the name of the author. PhD program in Network Engineering Enhanced Mobility Management Mechanisms for 5G Networks Doctoral thesis by: Akshay Jain Thesis Advisors: Dr. Elena López-Aguilera Dr. Ilker Demirkol Department of Network Engineering Barcelona, Spain July 2020 Copyright c  Akshay Jain Universitat Politècnica de Catalunya, BarcelonaTECH Acknowledgments This thesis is not just an embodiment of the work that has been done over the course of last three and half years, but it is also a symbol of how many people come together to make something, that at many stages seemed outright impossible, possible. At the very outset, I would like to express my profound gratitude to my doctoral thesis advisors Dr. Elena López-Aguilera and Dr. Ilker Demirkol. They have been instrumental in helping me develop not just my technical skills but also as a human being during this entire process. For me they have been like a family away from my family, and during the numerous coffees and work related trips I was able to learn extensively about the art of doing research. In a more formal setting, they have been instrumental in guiding me through some of the most difficult moments during my research by providing not just technical guidance but also moral support. For that I am eternally grateful to them. With these experiences, it is my hope that someday I will be able to emulate not just their wisdom but their kindness and empathetic nature as well, whilst forging my career in this increasingly competitive world. All of this would of course have not been possible without the unwavering support of my mother and father. They have forever been my bedrocks, and I truly believe that it is the values that they instilled in me which have allowed me to conquer many of the challenges that I met along this journey. Moreover, it is their sheer strength and patience which has inspired me to forge ahead at many junctures of my short career. For this I am indebted to them for life. I would also like to thank the rest of the family here, for being understanding and supportive whenever I needed them to be. I would like to specially mention Mr. Amit Jain for inspiring me to take up wireless communications and Ms. Pranjali Jain for being the elder sibling I never had and providing all the moral support. During the period of this thesis, I met many researchers who became colleagues, and ultimately best friends and in some cases even brothers/sisters. I would specially like to mention Mr. Rakibul Islam Rony, Mr. Girma M. Yilma, Mr. Mikel Irazabal, Dr. Francesco Devoti, Mr. Victor Baños-Gonzalez, Dr. Jorge E. Gaitán Pitre, Mr. Matteo Vincenzi, Mr. Nikolaos Giatsoglou, Mr. Lanfranco Zanzi, Dr. Jian Song, Dr. Alejandro S. Gonzalez, Dr. Adriana Fernández-Fernández, Dr. Leonardo Ochoa-Aday, Ms. Irian Leyva-Pupo, Mr. iii iv Alejandro, Mr. Carlos P., Mr. Ahmed, Mr. Asif Habibi, Dr. Khalid, Dr. Xavier Costa, Dr. Vincenzo Sciancalepore, Dr. Marco Di Renzo, Dr. Birkan H. Yilmaz, Dr. Sergi Abadal, Dr. Albert Cabellos-Aparicio, Dr. Eduard Alarcón, Dr. Ali Sadeghian, Dr. Lim Deoksu and Mr. Haazy Haastrup for playing a significant role towards the completion of my thesis. For the various conversations (technical/personal) and memorable moments, I am eternally grateful. Outside of work, my life in Barcelona has been enriched by interactions with people from various walks of life. They have helped me settle down in this vibrant city and have been my support throughout. I would like to specially mention Ms. Esther Xalabarder, Mr. Marti Rodrigo, Ms. Marisol, Ms. Laura Vargas, Mr. Carlos, Ms. Jacqueline, Mr. Flavio Fernandez, Ms. Silvina Sanchez, Mr. C Anand Iyer, Ms. Debarati Shome, Mr. Dhaval Gadariya, Ms. Judith Murray, Mr. Alan Urquhart and Ms. Boglarka Nagy for their unwavering support and helping me through this process. I will forever be grateful to you. My journey to this point has its roots to my time back in the USA. It is here where I learnt my art under the most challenging circumstances. But again I was blessed by the grace of god, and I met some of the best friends and colleagues over here. I would like to specially mention Ms. Stuti Joshi, Dr. Akanksha Pandey, Mr. Rahul Bhatia, Dr. Abhilash Paneri, Mr. Subodh Chaturvedi, Mr. Ankit Gupta, Mr. Chris Blower and Mr. Akash Dhruv for their unwavering support, belief and unmatchable acts of kindness towards me. For their friendship and contributions towards my development as a person and as a researcher, I am extremely grateful. I would also like to take this opportunity to mention the people in India who have been instrumental in my development as a person, since the time of my Bachelors. Specifically, I would like to thank Mr. Victor Roy, Mr. Siddhant Dash, Mr. Amit Kumar, Ms. Sadhvi Aggarwal, Mr. Animesh Kumar, Mr. Bhavik Gattani, Mr. Vinayak Iyer, Ms. Shruti Mahajan, Ms. Kanika Patoria, Mr. Nimish Shah, Mr. Nishant Tilokani and Mr. Keshav Mathur. I am eternally grateful and indebted to you for your extremely vital and significant contributions towards my life as a professional and as a person. Last but not the least, I would like to thank UPC, the doctoral school, the entire Network Engineering department, the various cafeteria and administrative staff members, the many anonymous journal/conference reviewers as well as the thesis reviewers for their unflinching support, guidance and consideration towards my goal of completing this thesis in the best possible manner. I would like to state here that, I have tried my best to mention everyone who has been a part of this journey. However, if in any case somebody’s name doesn’t appear, then I sincerely and whole-heartedly request your pardon. Preface Wireless standards such as 2G, 3G, 4G and currently 5G, promise incremental improvements in the Quality of Experience (QoE) and Quality of Service (QoS) when compared to their predecessor technologies (e.g., 5G promises better QoE and QoS than 4G/3G/2G). The quantitative measures of QoE and QoS relate to improved throughput, reliability, etc., from the perspective of the user. These measures of QoE and QoS are tightly coupled with the type of applications (e.g., Emergency services will require low latency and high reliability, while broadband services will require high bandwidth with less stringent latency and reliability measures as compared to emergency services). Further, up until 4G, industry and to some extent academia, through 3GPP, IETF and ETSI, defined methods that served the networks infallibly. However, with the industry facing a significant downturn in their revenues, the prospect of integrating other business verticals as well as moving towards a more softwarized (and thus economical) network deployment approach has led to the advent of 5G and hence, a revolution. However, such revolutions, as we may know from our knowledge in history, involves significant transformations in each section of the community. Similarly, many mechanisms that served the legacy telecommunication networks for months, years or decades are now being identified as being non-usable or at best sub-optimal. The reason being, increased heterogeneity, complexity and density within the 5G networks as compared to any other legacy system. And so, one such class of mechanisms, which are also extremely critical for any wireless standard, are the Mobility Management (MM) mechanisms. Mobility Management mechanisms ensure the seamless connectivity and continuity of service for a user when it moves away from the geographic location where it initially got attached to the network. But, and as we have already indicated, the 5G network characteristics render the legacy MM approaches as being either non-usable or inefficient. Hence, in this thesis, we firstly explore the various mechanisms that have been employed or conceived to perform Mobility Management in legacy (2G/3G/4G) as well as 5G networks. Further, based on the 5G requirements as well as the initial discussions on Beyond 5G networks, we provision a novel qualitative gap analysis. We also define the persistent v vi challenges that exist with regards to MM mechanisms for 5G and beyond networks. Based on these challenges, we define the potential solutions and a novel framework for the 5G and beyond MM mechanisms. This novel framework specifies a complete stack of MM mechanisms at the access network, core network and at the extreme edge network (users/devices) level, that will help satisfy the requirements for the 5G and beyond MM mechanisms. Following this, and as part of the defined novel MM framework, we present a novel ondemand MM service strategy. This on-demand feature provisions the necessary reliability, scalability and flexibility to the MM mechanisms. These three characteristics, as we elaborate in more detail in the thesis, will be the pillars of future MM mechanisms. It is important to state here that such an on-demand framework will ensure that appropriate resources and mobility contexts are defined for users who will have heterogeneous mobility profiles, i.e., pedestrian, vehicular, high speed traffic, etc., along side applications with versatile QoS requirements in a network with multiple Radio Access Technologies, such as LTE, Wi-Fi, 5G, etc. Next, based on the novel MM framework for 5G and beyond mechanisms that we have defined in this thesis, we tackle the problem of core network signaling that occurs during MM in 5G/4G networks. A novel handover signaling mechanism has been developed, which eliminates unnecessary handshakes during the handover preparation phase as well as preserves the legacy data structures. This not only allows for ease of transition to future softwarized network architectures but also simultaneously leads to significant reduction in latency, processing cost and transmission cost of handover signaling. Note that, to perform our analysis we utilized data from Greek and Japanese network operators as well as from telecom vendors such as Cisco. A further enhancement of the aforementioned proposed handover signaling mechanism has also been provided, wherein a premonition of a handover failure is utilized to design the handover preparation phase signaling. This consequently results in additional performance gains as observed through our quantitative evaluation. We then perform a comparative analysis of the proposed strategy and the 3GPP handover signaling strategy on a network wide deployment scenario, wherein the performance gains through our proposed strategy are further highlighted. Lastly, a novel user association and resource allocation methodology, namely AURA-5G, has been proposed. The developed methodology addresses scenarios wherein applications with heterogeneous requirements, i.e., enhanced Mobile Broadband (eMBB) and massive Machine Type Communications (mMTC), are present simultaneously. Consequently, a first approach in literature, wherein a joint optimization process for performing the user association and resource allocation while being cognizant of heterogeneous application requirements, has been performed. Concretely, the methodology aims at not only assigning an AP to a vii user, but also aims at reserving the appropriate resources, i.e., bandwidth at the chosen APs, given the heterogeneous application requirements and other prevailing network constraints. As mentioned, we capture the peculiarities of this important mobility management process through the various constraints, such as backhaul requirements, dual connectivity options, available access resources, minimum rate requirements, etc., that we have imposed on a Mixed Integer Linear Program (MILP). The objective function of this established MILP problem is to maximize the total network throughput of the eMBB users, while satisfying the minimum requirements of the mMTC and eMBB users defined in a given scenario. Through numerical evaluations we show that our approach outperforms the baseline user association scenario in terms of achievable system throughput for all possible constraint combinations. The baseline scenario being, to attach the users to an AP with the best Signal to noise ratio towards them and then dividing the access resources equitably amongst all competing users at a given AP. Moreover, to ensure the applicability of the devised methodology, we have presented a system fairness analysis, as well as a novel fidelity and complexity analysis for the same. Notably, for the fidelity analysis, we analyze how well the system satisfies the latency and backhaul utilization constraints. Further, for the complexity analysis, we observe the time to converge to an optimal solution as well as the number of Monte Carlo trials in which our framework is able to determine such an optimal solution, i.e., the solvability of the MILP problem. An extension of this work has then been briefly summarized in the future works section, wherein we comment about the possibility of integrating the Ultra-reliable low latency communication (URLLC) services into the AURA-5G framework via a multi-objective optimization methodology. Given the aforementioned efforts, we believe that this thesis has significantly advanced the area of Mobility Management for 5G and beyond networks by provisioning methods and system concepts that address many of the important persistent challenges. This has been reinforced by the broad acceptance of our work into multiple globally recognized conferences, reputed journals as well as a patent application. List of Publications [J3] A. Jain, E. Lopez-Aguilera, and I. Demirkol, "User Association and Resource Allocation in 5G (AURA-5G): A Joint Optimization Framework", Submitted to Elsevier Computer Networks, pp. 1–35, 2020. (Area: Computer Science; Quartile: Q1 (13/53); IF: 3.03 (2018)) [PT1] A. Jain, E. Lopez-Aguilera, and I. Demirkol, "Handover Method and System for 5G Networks", WO 2019/229219 A1 (WIPO PCT), pp. 1-98, Dec. 2019. (Positive International Search Report) [J2] A. Jain, E. Lopez-Aguilera, and I. Demirkol, "Are Mobility Management Solutions Ready for 5G and Beyond?", Accepted in Elsevier Computer Communications, pp. 1–36, 2020. (Area: Telecommunications; Quartile: Q2 (37/88); IF: 2.816 (2018)) [J1] A. Jain, E. Lopez-Aguilera, and I. Demirkol, "Evolutionary 4G/5G Network Architecture Assisted Efficient Handover Signaling", IEEE Access, vol. 7, pp. 256–283, Dec. 2018. (Area: Telecommunications; Quartile: Q1 (19/88); IF: 4.098 (2018)) [C4] A. Jain, E. Lopez-Aguilera, and I. Demirkol, "Improved Handover Signaling for 5G Networks", IEEE 29th Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC) 2018, pp. 164–170, Sept. 2018. [C3] A. Jain, E. Lopez-Aguilera, and I. Demirkol, "Enhanced Handover Signaling through Integrated MME-SDN Controller Solution", IEEE 87th Vehicular Technology Conference VTC Spring 2018, pp. 1–7, Jun. 2018. [C2] R. I. Rony, A. Jain, E. Lopez-Aguilera, E. Garcia-Villegas, and I. Demirkol, "Joint access-backhaul perspective on mobility management in 5G networks", IEEE Conference on Standards for Communications and Networking, CSCN 2017, pp. 115–120, Sept. 2017. [C1] A. Jain, E. Lopez-Aguilera, and I. Demirkol, "Mobility Management as a Service for 5G Networks", IEEE ISWCS 2017, pp. 1–6, Jun. 2017. viii 5.6 Proposed Handover signaling sequence for Inter-RAT HO from 5G NGC to EPS......................................... 113 5.7 Proposed Handover cancel phase signaling for Inter-RAT HO from 5G NGC toEPS. ...................................... 115 5.8 Proposed Handover rejection phase signaling for Inter-RAT HO from LTE to 3G/2G network when there is a S-GW relocation and indirect tunneling exists.115 5.9 Handover failure aware Handover preparation Signaling for Inter-RAT HO from5GNGCtoEPS............................... 117 5.10 Optimal proposed Handover rejection phase signaling sequence for Inter-RAT HOfrom5GNGCtoEPS............................. 118 5.11 Handover failure aware Handover preparation Signaling for Inter-RAT HO from LTE-EPC to 3G/2G when there is indirect tunneling and S-GW relocationoccurs. .................................... 118 5.12 Optimal proposed Handover rejection phase signaling sequence for Inter-RAT HO from LTE-EPC to 3G/2G when there is indirect tunneling and S-GW relocationoccurs.................................. 118 5.13 Handover preparation scenario: Transmission cost analysis for the Japanese operator deployment (X-axis notations have been re-utilized from Tables 5.45.7). ........................................ 130 5.14 Handover preparation scenario: Transmission cost analysis for the Greek operator deployment (X-axis notations have been re-utilized from Tables 5.4-5.7).131 5.15 Handover failure scenario: Transmission cost analysis for the Japanese operator deployment (X-axis notations have been re-utilized from Tables 5.4-5.7). 132 5.16 Handover failure scenario: Transmission cost analysis for the Greek operator deployment (X-axis notations have been re-utilized from Tables 5.4-5.7). . . 133 5.17 Network wide processing cost analysis. . . . . . . . . . . . . . . . . . . . . . 140 5.18 Network wide occupation time analysis. . . . . . . . . . . . . . . . . . . . . . 141 5.19 Proposed evolutionary network architecture. . . . . . . . . . . . . . . . . . . 142 5.20 SDN agent for the evolutionary network architecture. . . . . . . . . . . . . . 145 5.21 SDN-enabled Mobility Management unit (SeMMu) architectural framework. 147 6.1 AURA-5G Framework. The logical flow, i.e. flow of control, within the developed tool is depicted using dashed arrows, whilst solid arrows indicate the dataflowintheprogram. ............................ 154 6.2 Illustrative example of the network topology under study . . . . . . . . . . . 169 xv 6.3 Total Network Throughput for multiple combination of constraints being employed on (a) CABE, (b) CMBE, (c) CAIE and(d) CMIE scenarios. . . . . . 174 6.4 Total Network Throughput for multiple combination of constraints being employed on (a) SABE, (b) SMBE, (c) SAIE and (d) SMIE scenarios. . . . . . 175 6.5 Circular and Square deployment characteristics for SCs around MCs. . . . . 176 6.6 Total Network Throughput for multiple combination of constraints being employed on (a) CABE, (b) CABEm, (c) CAIE and (d) CAIEm scenarios. . . . 177 6.7 Total Network Throughput for multiple combination of constraints being employed on (a) CMBE, (b) CMBEm, (c) SAIE and (d) SAIEm scenarios. . . . 178 6.8 Jain’s Fairness index deviation measure for user throughputs over multiple combination of constraints being employed on (a) CABE, (b) CMBE, (c) CAIE and (d) CMIE scenarios. . . . . . . . . . . . . . . . . . . . . . . . . . 181 6.9 Jain’s Fairness index deviation measure for multiple combination of constraints being employed on (a) SABE, (b) SMBE, (c) SAIE and (d) SMIE scenarios. ..................................... 183 6.10 Jain’s Fairness measure for multiple combination of constraints being employed on (a) CABE, (b) CABEm, (c) CAIE and (d) CAIEm scenarios. . . . 184 6.11 User Throughput Distribution for Dual Connectivity (DC) with Minimum Rate (MRT) constraints in (a) CEBAS, (b) CEBMS, (c) CEIAS, (d) CEIMS, (e) SEBAS and (f) SEBMS scenarios. . . . . . . . . . . . . . . . . . . . . . . 186 6.12 User Throughput Distribution for Dual Connectivity (DC) with Minimum Rate (MRT) constraints in (a) SEIAS and (b) SEIMS scenarios. . . . . . . . 187 6.13 Backhaul Utilization for Dual Connectivity (DC) and DC with Backhaul Capacity constraints in (a) CABE, (b) CMBE, (c) CAIE, (d) CMIE, (e) SABE and (f) SAIE scenarios. Red colored BS indices are for MCs and the rest for SCs. ........................................ 189 6.14 Backhaul Utilization for Dual Connectivity (DC) and DC with Backhaul Capacity constraints in (a) CABEm and (b) CAIEm scenarios. Red colored BS indices are for MCs and the rest for SCs. . . . . . . . . . . . . . . . . . . . . 190 6.15 Observed Latency for (a) CABE, (b) CMBE, (c) SABE, (d) SMBE, (e) CAIEm and (f) CMIEm scenarios. . . . . . . . . . . . . . . . . . . . . . . . 192 6.16 Convergence time CDF (Empirical) for (a) CABE, (b) CAIE, (c) CMBE and (d)CMIEscenarios. ............................... 194 6.17 Convergence time CDF (Empirical) for (a) SABE, (b) SAIE, (c) CAIEm and (d)CMIEmscenarios. .............................. 195 xvi 6.18 Optimizer Status for (a) CABE, (b) CAIE, (c) CMBE, (d) CMIE, (e) SABE and (f) SAIE scenarios with 275 eMBB users. . . . . . . . . . . . . . . . . . 197 6.19 Optimizer Status for (a) CAIEm and (b) CMIEm scenarios with 275 eMBB users......................................... 198 6.20 Optimizer Status for (a) SABEm without Relaxed Backhaul, and (b) SABEm with Relaxed Backhaul scenarios with 275 eMBB users. . . . . . . . . . . . . 200 6.21 System Fairness Measure for (a) SABEm without Relaxed Backhaul, and (b) SABEm with Relaxed Backhaul scenarios with 275 eMBB users. . . . . . . . 200 6.22 Total Network Throughput for (a) SABEm without Relaxed Backhaul, and (b) SABEm with Relaxed Backhaul scenarios with 275 eMBB users. . . . . . 201 6.23 Optimizer Status for (a) SABEm with Relaxed Backhaul and Increased SC density scenario with 275 eMBB users, and (b) SABEm scenario with Relaxed Backhaul, Increased SC density, 5 ms downlink latency requirement and 275 eMBBusers..................................... 201 6.24 System Fairness Measure for (a) SABEm with Relaxed Backhaul and Increased SC density scenario with 275 eMBB users, and (b) SABEm scenario with Relaxed Backhaul, Increased SC density, 5 ms downlink latency requirementand275eMBBusers............................. 202 6.25 Total Network Throughput for (a) SABEm with Relaxed Backhaul and Increased SC density scenario with 275 eMBB users, and (b) SABEm scenario with Relaxed Backhaul, Increased SC density, 5 ms downlink latency requirementand275eMBBusers............................. 202 6.26 Optimizer Status for (a) CABEm without Relaxed Backhaul, and (b) CABEm with Relaxed Backhaul scenarios with 275 eMBB users. . . . . . . . . . . . . 203 6.27 System Fairness Measure for (a) CABEm without Relaxed Backhaul, and (b) CABEm with Relaxed Backhaul scenarios with 275 eMBB users. . . . . . . . 203 6.28 Total Network Throughput for (a) CABEm without Relaxed Backhaul, and (b) CABEm with Relaxed Backhaul scenarios with 275 eMBB users. . . . . . 205 6.29 Optimizer Status for (a) CABEm with Relaxed Backhaul and Increased SC density scenario with 275 eMBB users, and (b) CABEm scenario with Relaxed Backhaul, Increased SC density, 5 ms downlink latency requirement and 275 eMBBusers..................................... 205 6.30 System Fairness Measure for (a) CABEm with Relaxed Backhaul and Increased SC density scenario with 275 eMBB users, and (b) CABEm scenario with Relaxed Backhaul, Increased SC density, 5 ms downlink latency requirementand275eMBBusers............................. 206 xvii 6.31 Total Network Throughput for (a) CABEm with Relaxed Backhaul and Increased SC density scenario with 275 eMBB users, and (b) CABEm scenario with Relaxed Backhaul, Increased SC density, 5 ms downlink latency requirementand275eMBBusers............................. 207 A.1 Proposed Handover Signaling for LTE to 3G/2G Inter-RAT HO with Target S-GWandDirectTunnel. ............................ 218 A.2 Proposed Handover Signal mapping for LTE to 3G/2G Inter-RAT HO with Target S-GW and Direct Tunnel. . . . . . . . . . . . . . . . . . . . . . . . . 219 A.3 Proposed Handover Signaling for LTE to 3G/2G Inter-RAT HO without Target S-GW and Direct Tunnel. . . . . . . . . . . . . . . . . . . . . . . . . . . 220 A.4 Proposed Handover Signal mapping for LTE to 3G/2G Inter-RAT HO without Target S-GW and Direct Tunnel. . . . . . . . . . . . . . . . . . . . . . . . . 221 A.5 Proposed Handover Signaling for LTE to 3G/2G Inter-RAT HO without Target S-GW and Indirect Tunnel. . . . . . . . . . . . . . . . . . . . . . . . . . 222 A.6 Proposed Handover Signal mapping for LTE to 3G/2G Inter-RAT HO without Target S-GW and Indirect Tunnel. . . . . . . . . . . . . . . . . . . . . . . . 223 A.7 Proposed Handover Signaling for LTE to 3G/2G Inter-RAT HO with Target S-GW and Indirect Tunnel. . . . . . . . . . . . . . . . . . . . . . . . . . . . 224 A.8 Proposed Handover Signal mapping for LTE to 3G/2G Inter-RAT HO with Target S-GW and Indirect Tunnel. . . . . . . . . . . . . . . . . . . . . . . . 225 A.9 Proposed Handover Signaling for 3G/2G to LTE Inter-RAT HO without TargetS-GW...................................... 226 A.10 Proposed Handover Signal mapping for 3G/2G to LTE Inter-RAT HO without TargetS-GW.................................... 227 A.11 Proposed Handover Signaling for 3G/2G to LTE Inter-RAT HO with Target S-GW........................................ 228 A.12 Proposed Handover Signal mapping for 3G/2G to LTE Inter-RAT HO with TargetS-GW.. .................................. 229 A.13 Proposed Handover Signaling for LTE Intra-RAT HO with Target S-GW and MME. ....................................... 230 A.14 Proposed Handover Signal mapping for LTE Intra-RAT HO with MME relocation (without S-GW relocation). . . . . . . . . . . . . . . . . . . . . . . . . 231 A.15 Proposed Handover Signaling for LTE to 3G/2G Inter-RAT HO without Target SGW and Direct Tunnel. . . . . . . . . . . . . . . . . . . . . . . . . . . . 232 xviii A.16 Proposed Handover Signal mapping for LTE to 3G/2G Inter-RAT HO without Target SGW and Direct Tunnel. . . . . . . . . . . . . . . . . . . . . . . . . . 233 A.17 Proposed Handover Signaling 5G Inter NG-RAN N2 based Handover. . . . . 234 A.18 Proposed Signaling for 5G core to EPS Handover with N26 Interface. . . . . 235 A.19 Proposed Signaling for EPS to 5G Core Handover with N26 Interface. . . . . 236 A.20 Proposed Signaling for EPS to 5G Core Handover Cancel. . . . . . . . . . . 237 A.21 Proposed Signaling for 5G Core to EPS Handover Cancel. . . . . . . . . . . 238 A.22 Proposed Signaling for EPS to 5G Core Handover without N26 interface: PDU establishment.................................... 239 A.23 Proposed Signaling for 5G Core to EPS Handover without N26 interface: UE requestedConnectivity............................... 240 xix List of Tables 1.1 Expectations from 5G Networks . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.1 Functional Requirements from 5G and beyond MM . . . . . . . . . . . . . . 17 3.1 Governing Parameters for the Reliability, Scalability and Flexibility of a MM mechanism/standard ............................... 57 3.2 Compliance with the Reliability, Scalability and Flexibility criteria for the legacy MM mechanism/standard . . . . . . . . . . . . . . . . . . . . . . . . 65 3.3 Compliance with Reliability, Scalability and Flexibility criteria of Current state-of-the-art MM mechanism/standard . . . . . . . . . . . . . . . . . . . . 77 3.4 Mapping potential solutions to MM challenges . . . . . . . . . . . . . . . . . 84 4.1 Comparison between MMaaS and current/legacy architecture . . . . . . . . . 93 5.1 Different handover scenarios analyzed . . . . . . . . . . . . . . . . . . . . . . 109 5.2 Link Type and Corresponding Delays in Proposed Architecture (Derived from a Japanese Operator [176] and Cisco data [177] . . . . . . . . . . . . . . . . 123 5.3 Link Type and Corresponding Delays in Proposed Architecture (Derived from a Greek Operator and Cisco data [177]) . . . . . . . . . . . . . . . . . . . . . 124 5.4 Preparation Phase: Handover Latency Improvement Analysis (Cisco and CellularOperator-Japan) .............................. 126 5.5 Preparation Phase: Handover Latency Improvement Analysis (Cisco and CellularOperator-Greece) .............................. 127 5.6 Failure Phase: Handover Latency Improvement Analysis (Cisco and Cellular Operator-Japan) ................................. 128 5.7 Failure Phase: Handover Latency Improvement Analysis (Cisco and Cellular Operator-Greece)................................. 128 5.8 Processing Cost Analysis for Handover Preparation phase . . . . . . . . . . . 134 5.9 Processing Cost Analysis for Handover Failure phase . . . . . . . . . . . . . 134 xx 5.10 Handover failure aware signaling design analysis . . . . . . . . . . . . . . . . 136 5.11 Message size Computation: Inter-RAT HO from LTE to 3G/2G when S-GW is relocated and indirect tunneling exists . . . . . . . . . . . . . . . . . . . . 138 5.12Messagesizeanalysis............................... 139 6.1 Definitions list for Notations, Variables and Constants . . . . . . . . . . . . . 158 6.2 AnalyzedScenarios................................ 164 6.3 Constraint Combinations for Scenarios . . . . . . . . . . . . . . . . . . . . . 168 6.4 EvaluationParameters .............................. 170 xxi List of Abbreviations 2G Second generation 3G Third generation 3GPP Third Generation Partnership Project 4G Fourth generation 5G Fifth generation 5G NORMA 5G Novel Radio Multiservice adaptive network Architecture 5G NR 5G New Radio 5GPPP The Fifth Generation infrastructure Public Private Partnership ANDSF Access Network Discovery and Selection Function AMF Access and Mobility management Function AR Access Router AuR Augmented Reality ASN.1 Abstract Syntax Notation number one BBU Baseband Unit BCE Binding Cache Entry BH Backhaul BS Base Station CA Carrier Aggregation CAPEX Capital Expenditure CDN Content Delivery Network CN Core Network CoA Care-of-Address CoMP Coordinated Multi-Point transmission CP Control Plane CrN Correspondent Node CRAN Centralized Radio Access Network CSI Channel State Information D2D Device-to-Device xxii DL Downlink DP Data Plane E2E End-to-end EDGE Enhanced Data Rate for GSM Evolution eNB Evolved Node-B EPC Evolved Packet Core ETSI European Telecommunications Standards Institute E-UTRAN Evolved UMTS Terrestrial Radio Access Network FA Forwarding Agent F-AP Fully equipped Access Points FBAck Fast Binding Acknowledgement FBU Fast Binding Update FDD Frequency Division Duplex FH Fronthaul FHO Frequent Handover FMIPv6 Fast MIPv6 FNA Fast Neighbor Acknowledgement FP7 7th Framework Programme for Research and Technological Development GA Genetic Algorithm gNB next generation NodeB GPRS General Packet Radio Services GSM Global System for Mobile Communications HA Home Address HARQ Hybrid Automatic Repeat Request HetNet Heterogeneous Networks HMIPv6 Hierarchical MIPv6 HO Handover HSPA High Speed Packet Access IEEE Institute of Electrical and Electronics Engineers IE Information Element ITU-R International Telecommunication Union-Radio Communication Section KPI Key Performance Indicator LCoA Local CoA LIPA Local IP Access LMA Local Mobility Anchor LOS Line of Sight xxiii LTE Long Term Evolution LTE-A Long Term Evolution Advanced LWA LTE-WLAN Aggregation MAC Medium Access Control MADM Multi-Attribute Decision Making MAG Mobility Access Gateway MAP Mobility Anchor Point MBS Macro-Base Station MC Macro-cell MIH Media Independent Handover MIMO Multiple Input and Multiple Output MIPv4 Mobile IPv4 MIPv6 Mobile IPv6 MME Mobility Management Entity MMT Multi-Mode Terminal mmWave Millimetre Wave MN Mobile Node MPTCP Multipath Transmission Control Protocol MTC Machine Type Communication MU-MIMO Multi-User MIMO NBI Northbound Interface NC Network Controller NFV Network Function Virtualization NFVO Network Function Virtualization Orchestrator NGC Next Generation Core NGFI Next Generation Fronthaul Interface NGPON Next Generation Passive Optical Network NG-RAN Next Generation Radio Access Network NLoS Non-Line of Sight OF OpenFlow OFDM Orthogonal Frequency Division Multiplexing OFDMA Orthogonal Frequency Division Multiple Access OPEX Operating Expenditure PCRF Policy and Charging Rules Function PDCP Packet Data Convergence Protocol P-GW Packet Gateway xxiv CHAPTER 1. INTRODUCTION 5 reduction in service times as well as provisioning of compute facilities near the network edge, is the MEC paradigm [19]. Furthermore, the D2D communications will enable information sharing as well as extended connectivity near the network edge [7]. This will facilitate 5G networks in provisioning better scenario specific, i.e., context-based, services. Additionally, the UDN, M-RAT, mmWave and Joint access and backhaul strategies, aim at provisioning extreme flexibility on the radio side in terms of resource availability as well as resource sharing. Specifically, UDN and mmWave techniques aim at increasing the spectral efficiency of the network by bringing the BSs closer to the users and opening up the higher frequency bands, respectively [7]. Moreover, via the M-RAT technique, it will be possible for any given user to be able to connect to multiple BSs belonging to different RATs. Through Release-15, 3GPP has already standardized the concept and functional characteristics of dual connectivity [20]. And whilst the aforementioned strategies primarily increase the resource availability, the Joint access and backhaul design mechanism aims to address the issue of on-demand and context based resource sharing. Multiple works, such as [C2], have already envisioned how the joint design mechanism can enhance the performance of 5G networks. Lastly, C-RAN aims at provisioning a flexible RAN deployment procedure, and hence, a flexible RAN split. This essentially will assist operators in deploying lower cost RRHs and centralizing the processing of the data, which will eventually lead to significant processing gains [21]. However, and also according to our contribution [J2], these aforesaid enablers do not instill the required reliability, flexibility and scalability necessary to ensure the seamless mobility aspect of 5G networks. This is so because, while the SDN and NFV paradigms give a global view of the network, the signaling required to gather such information for mobile users can quickly drown the entire network with control messages [J2]. The DMM paradigm on the other hand, whilst handling the mobility without a central controller and solving the core network signaling and latency issue [22–24], can be quite detrimental at the access network level. The reason being that it requires control signaling amongst the routers to exchange the context of the migrating user. Any disruption in the link or an abruptly large number of migrations can present significant challenges to the DMM strategy. Additionally, techniques such as M-RAT, UDN, C-RAN and mmWave, complicate the development of an effective MM strategy because they increase the dimensionality of the search space as well as alter the behavior of the physical channel, as compared to the sub-6 GHz based 2G/3G/4G standards. Lastly, the MEC paradigm through its close proximity to the access network can help alleviate issues regarding latency as well as core network signaling. However, when users are mobile their services will also need to be replicated/migrated. So far research efforts such CHAPTER 1. INTRODUCTION 6 as [25–27] have not been able to provision methods which meet the latency and efficient compute capacity utilization requirements in the event that service replication/migration is required. 1.2 Motivation From our discussions so far, it is evident that in 5G networks the design and development of MM solutions will be faced with multitude of challenges. It is these existent challenges that have contributed significantly to the motivation for the work that has been presented in this thesis. Hence, we firstly consolidate these challenges as follows: •The ultra high density characteristic of the 5G networks, wherein there will be an exponential increase in the number of devices it serves, will be an important challenge [7, 9]. Moreover, a similar increase in the number of BSs, which will cater to these devices, is also expected. Thus, to manage the mobility contexts as well as the signaling involved in such a dense scenario will present a significant challenge. •Extreme heterogeneity in the network, wherein the users have different services with different QoS requirements along side the heterogeneous RATs provided by network operators, will pose significant challenges towards the design and development of future MM solutions. The reason being that, currently a one size fits all approach is being utilized. However, given the aforesaid heterogeneity, it will be important that the future MM strategies consider each service type’s requirement individually. Note that, here by different services we mean the Ultra-reliable low latency (URLLC), enhanced Mobile Broadband (eMBB) and massive Machine Type Communications (mMTC) services, as defined in the 5G standards [7]. Concretely, the URLLC services will require extreme reliability, not only in terms of the bit error rate performance but also in terms of link outage probability, as well as low latency, approximately 1ms or less. Furthermore, the eMBB services will necessitate extremely high data rates, i.e., upto 10 Gbps, to support applications such as VR/AuR, etc. Additionally, the mMTC services will require that the network supports extremely large deployment of devices, i.e., of the order of 106 per km2[7,9]. •Given the heterogeneity in service types, RAT types, as well as the broader range of support, in terms of speeds (upto 500 km/hr), that the 5G network aims to support, the users will consequently have a broader variety of mobility profiles. Furthermore, the increased density of network and the choice of RATs will translate to an extremely high CHAPTER 1. INTRODUCTION 7 dimensional solution space to determine appropriate user-BS associations and resource allocation schemes. Thus, to be able to determine these appropriate resources and associations, future MM schemes will have to traverse through the aforesaid extremely high dimensional solution space to find an optimal solution. This will consequently perpetuate the challenging nature of 5G scenarios for MM. •The current methods as well as legacy mechanisms do not provision a unified and complete MM framework. Concretely, multiple studies highlighting individual MM mechanisms, such as DMM [22–24], LTE handover [12], Dual Connectivity [20], etc., exist. However, as stated above, a comprehensive suite of MM strategies at the access, core and extreme edge (users/devices) level, that will help satisfy the 5G MM requirements, is not available yet. Given these broad challenges, the motivation for the work done in this thesis is to build on the decades of experience gained by the research community in developing effective MM strategies, and advance it such that the newly developed MM strategies can cater to the requirements of the 5G networks and even beyond. Such an approach should also compulsorily take into account the 5G enablers mentioned in Section 1.1.2. Further, the current 5G standards by 3GPP [28,29] provision mobility management methods through the handover signaling sequences as specified in [29] as well as procedures for communicating with non-3GPP techniques through LAA and LWA. Additionally, the academic community has presented multiple solutions for 5G MM. However, specific applicable and implementable MM procedures that will be able to handle the complexity, which the 5G framework introduces, continue to elude. Consequently, another motivating factor for the work done in this thesis has been to also develop solutions that are tangible, adaptable and employable by both industry and academia. And so, in the following section we highlight the objectives of this thesis, followed by the contributions of the work embodied in this thesis to achieve the aforementioned objectives. 1.3 Objectives and Contributions While the main objective of this thesis is to investigate and develop Enhanced Mobility Management mechanisms for 5G Networks, we broke it down further into several relatively smaller yet significantly important and critical objectives. We discuss them as follows: •Suitability of existing MM algorithms: While we seek to innovate and develop new methods to handle the complex network scenario that 5G will present, it is always CHAPTER 1. INTRODUCTION 8 prudent to understand and analyze if already existing mechanisms can assist in meeting these requirements in part or in full. Henceforth, we developed a novel qualitative analysis whereby we investigated some of the most prevalent MM approaches from the legacy mechanisms as well the MM mechanisms being developed for 5G networks by current academic and industrial efforts. Additionally, we have also taken cognizance of the emerging discussions related to Beyond 5G (B5G) networks and their corresponding enablers. As a consequence, we have extended this particular study to B5G scenarios as well. And so, through the aforementioned qualitative analysis we have concretized the persistent challenges that continue to exist, potential solutions to these challenges and a framework for future MM mechanisms, thus also paving the way for our subsequent research efforts. An embodiment of this work, i.e., reference [J2], titled “Are mobility management mechanisms ready for 5G and Beyond?” has been submitted to Elsevier Computer Communications Journal and is currently under review. •On-demand Mobility Management: An important aspect for the MM mechanisms to cater the extremely dense and heterogeneous 5G networks will be to provision a flexible and on-demand strategy. This is so because, the current day networks provision a one size fits all approach. Such a strategy will be counter-productive given the 5G network characteristics. Hence through our work, titled “Mobility Management as a Service for 5G Networks” [C1], which was presented at the 2017 IEEE ISWCS conference, we have provisioned an on-demand MM framework which also explores the multiple avenues of introducing flexibility in provisioning MM service. •Handover mechanisms for 5G networks: One of the most basic elements of managing mobility of users is to be able to allow them to seamlessly transition from one BS to another. This seamless transition can be either in the same tracking area/Central entity domain/PLMN or they might be transitioning to an BS of another tracking area/Central entity domain/PLMN. By central entity here we mean SGSN/GGSN for 2G/3G networks, MME in 4G-LTE and SMF in the 5G network. In any of the aforementioned scenarios, continuity of service with the required QoS needs to be maintained. This burdens the current handover strategies because 5G networks will be extremely dense and heterogeneous. Hence, we developed a novel handover method and system that enhances the current handover mechanisms by upto 50% in terms of latency, processing cost and transmission cost for the various handover scenarios defined by 3GPP [29]. Note, the evaluations were performed by utilizing real network data from Greek and Japanese operators as well as from vendors such as Cisco. Further, an SDN based method and system has been proposed which not only enhances 5G CHAPTER 1. INTRODUCTION 9 handover signaling, but it also enhances the inter-RAT handover between 5G/4G and 4G/3G-2G networks. Consequently, some of the initial work that focused on the 4G Intra-RAT and 4G/3G-2G Inter-RAT handover method was published as “Enhanced handover signaling through integrated MME-SDN controller solution” in 2018 IEEE 87th Vehicular Technology Conference (VTC Spring) [C3]. Further, a subset of the 5G/4G InterRAT Handover signaling method wherein the N26 interface does not exist was then presented as “Improved handover signaling for 5G networks" in 2018 IEEE 29th Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC) Conference [C4]. This work was then followed by a publication titled “Evolutionary 4G/5G network architecture assisted efficient handover signaling" in IEEE Access Journal [J1]. The work focused on other 5G Interand Intra-RAT handovers as well as a novel handover preparation aware handover rejection methodology. Further, a novel network wide analysis, utilizing data from a Greek and Japanese network operator, was also provided. Lastly, based on the work done with regards to the HO signaling along side a more detailed development of the SDN based system, a patent application titled “Handover Systems and Method for 5G Networks” was filed with the European Patent Office [PT1]. Currently, we have obtained a positive international search report (ISR) for our PCT application. •Application aware User Association Methods: In 5G, the heterogeneity will arise not only from the different type of BSs but also from the different application types that will need to be served. Thus, in order to ensure that they receive the requested QoS, it will be equally critical to determine the best application to BS association. In our work we consider only single application per user, and hence, it can be treated as a traditional user association paradigm. However, note that the work presented in this thesis can be easily extended to multiple applications, with different QoS requirements, per user. Henceforth, we developed a novel Mixed Integer Linear Programming (MILP) based optimization framework, known as AURA-5G, that evaluates the topology and finds the most optimal application (user) association given the multiple real network constraints. Further, we have evaluated scenarios wherein only eMBB services exist, and where both eMBB and mMTC services exist together. Through this work we have also provisioned a working tool for the community so that it can be utilized for research as well as implementation. Given, implementation being one of the intended aspects, we have additionally performed extensive fidelity, performance and complexity analysis. As CHAPTER 1. INTRODUCTION 10 a consequence, an embodiment of this work, titled “User Association and Resource Allocation in 5G (AURA-5G): A Joint Optimization Framework”, is currently under review with the Elsevier Computer Networks journal [J3]. 1.4 Thesis Outline With the motivation and contributions of our work now highlighted, we specify the organization of this thesis in the text that follows. In Chapter 2, we discuss the state of the art of mobility management strategies as well as the various avenues where it can be employed. We also take cognizance of the emerging studies related to B5G networks and its enablers from the perspective of mobility management. Next in Chapter 3, a novel qualitative gap analysis, wherein we have evaluated the legacy as well as the currently proposed MM mechanisms, has been provided. Following this, we have presented a novel discussion on the persistent challenges, potential solutions to these challenges and a framework for 5G and beyond MM mechanisms. This consequently lays down the foundation for our subsequent work and also chapters. Thus, in Chapter 4, we present a novel on-demand MM paradigm. We detail its concept and methodology as well as the various benefits it presents for 5G MM mechanisms. Further in Chapter 5, we present an extensive discussion on handover signaling and the current 5G standards for the same. We then highlight the shortcomings and present our approach. Following this discussion, we present our analytical approach and the resultant comparative analysis for the myriad scenarios that 3GPP specifies. Additionally, a novel network wide analysis has also been presented to concretize the benefits that our approach provisions for 5G networks over the current standards. Then, in Chapter 6, we explore another dimension of MM methods wherein a new novel user association strategy has been proposed. We also propose a novel framework, namely AURA-5G, which can be utilized/implemented by industry and academia. Henceforth, we also present an analysis for the framework highlighting its fidelity and complexity. Lastly, we provide conclusions for the work done in this thesis in Chapter 7 and then discuss our future work proposals in Chapter 8. The thesis is concluded with an Appendix that consists of additional figures that have not been illustrated in the main text. Chapter 2 State of the Art in Mobility Management Overview In this chapter we present a detailed background with regards to the main mobility management mechanisms conceived and developed for the wireless networks. We firstly highlight the various functional requirements from future MM procedures, which is then followed by a detailed discussion on the various mechanisms/strategies in mobility management. These mechanisms/strategies are categorized as legacy and current state-of-the-art mechanisms/strategies. Note that, these discussions are carried out while being cognizant of the ongoing discussions with regards to B5G networks and their enablers. Additionally, we provision a novel 5G Service based architecture diagram along side a unique classification of the current state-of-the-art mechanisms. Note that, in Chapter 3, we build upon this state of the art and present a novel qualitative analysis with regards to the suitability of the myriad MM mechanisms, discussed in this chapter, for 5G and beyond MM mechanisms. Contributions [J2] A. Jain, E. Lopez-Aguilera, and I. Demirkol, "Are Mobility Management Solutions Ready for 5G and Beyond?", Accepted in Elsevier Computer Communications, pp. 1–36, 2020. (Quartile: Q2; IF: 2.816 (2019)) [C2] R. I. Rony, A. Jain, E. Lopez-Aguilera, E. Garcia-Villegas, and I. Demirkol, "Joint access-backhaul perspective on mobility management in 5G networks", IEEE Conference on Standards for Communications and Networking, CSCN 2017, pp. 115–120 Future wireless networks define a very challenging environment for mobility management (MM) solutions, due to the significant increase in density (in terms of both users and de11 CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 12 ployed base stations), in heterogeneity (given the various radio access technologies (RATs) supported), as well as in programmability (the network as well as the environment can be programmable). To achieve an ubiquitous network service in such challenging environments, it is critical to devise effective MM strategies that facilitate seamless mobility by allowing users to traverse through the network without losing connectivity and service continuity. One of the traditional approaches for allowing applications to serve a user in mobile scenarios has been to maintain network connectivity through handovers based on criteria such as Radio Signal Strength Indicator (RSSI), Signal to Interference and Noise Ratio (SINR), Reference Signal Received Quality (RSRQ), Reference Signal Received Power (RSRP), etc. However, in addition to the signal quality parameter centric handovers, modern day applications necessitate that other parameters such as available core network bandwidth, Endto-End (E2E) latency, backhaul bandwidth and backhaul reliability [30] are also taken into consideration. Moreover, maintaining Quality of Service (QoS), e.g., provisioning service continuity, link continuity, required bit-rate and latency, during mobility scenarios has been one of the primary objectives for novel MM mechanisms. Multiple strategies to satisfy such QoS criteria such as service migration [31], service replication [26], path reconfiguration [24], etc., have been proposed by the research community. MM solutions for 5G and beyond networks are also expected to ensure E2E connectivity and session continuity through the maintenance/preservation of IP address of the user towards the core network entity that provisions the service for the corresponding user. To motivate further, we consider an illustrative example of the future mobility scenario is presented in Figure 2.1, which shows the extraordinary nature of complexity that the future networks will present for MM. As shown in the Figure 2.1(a), a mobile user equipment (UE) is connected to multiple RATs (5G BS/ Long Term Evolution (LTE) evolved NodeB (eNB)/visible light communications (VLC) and Light Fidelity (LiFi) Small-cells [32–34], etc.), while having a delay tolerant and a delay sensitive application datastream (flows) with distinct QoS profiles. Also, the BS through which the delay tolerant flow is being served to the user has a good wireless link with a meta-surface in the vicinity. Note that, meta-surfaces are thin, but electrically significant, surfaces that enable the possibility of engineering the channel through the manipulation of phase, amplitude and polarization of the incident wave [35–37]. In addition to the meta-surfaces, future networks will also consist of mobile BSs such as drones, as shown in Figure 2.1(a). Note that, the density of meta-surfaces and drone BSs will also be extremely high in future networks. Further, in the scenario illustrated, we consider the use case wherein the drone BS is servicing a D2D cluster, and connecting it to the core network through one of the ground based BSs. The D2D cluster over the course of its existence does not generate packets as frequently as the CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 13 Device-to-Device Cluster UE Core Network Heterogeneous Access Network Device-to-Device Cluster Device-to-Device Cluster Core Network Heterogeneous Access Network UE Device-to-Device Cluster Core Network Heterogeneous Access Network UE UE Heterogeneous Access Network Core Network 2G/3G Base Stations Good wireless link LTE-NB/5G Base Stations LTE/VLC/ LiFi small cell Wi-Fi Access Point Core Network Element IoT Devices IP Core Prospective good wireless link Device-to-Device cluster flow New delay sensitive flow Delay sensitive flow Delay tolerant flow Degrading wireless link (a) (b) (c) (d) Meta-Surface Drone Base Station Figure 2.1: An illustrative 5G and beyond network mobility scenario. CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 14 other users, since the cluster devices mainly host Internet of Things (IoT) applications. Next, in Figure 2.1(b), as the user moves, it starts to register wireless links with better signal quality from other BSs as compared to those it is already associated to. It is imperative to state here that, the BSs can be from the same or different network operators. Henceforth, a careful and efficient RAT and BS selection for each flow will be necessary as part of the future MM mechanisms. It is interesting to observe that while the BS used for serving the delay tolerant flow in Figure 2.1(a) no longer has a good link quality, through the metasurfaces and their programmable nature it still has a good wireless link to the user and hence is able to serve it. Following the new RAT/BS association, flows pertaining to the user are redirected through the most optimal path. Novel MM mechanisms that aim to service the 5G and B5G networks will require efficient route optimization methods to perform the same. Additionally, the MM mechanisms will also need to implement IP forwarding so as to ensure E2E link continuity. In Figure 2.1(c) we then observe that as the user moves further, the RAT/BS selection and optimal routing methods are continually implemented. Further, when a new application request is generated, as seen in Figure 2.1(c), an appropriate RAT and BS for the given flow is selected alongside the route that satisfies the requested QoS. Lastly, in Figure 2.1(d), it can be seen that alongside the user’s flows, the D2D cluster’s flows are also being serviced by network. However, the D2D cluster is firstly serviced by a drone BS, which then relays information to/from the ground based BSs. These ground based BSs assist in serving the data flows generated from the devices in the D2D cluster by relaying the data to the relevant servers in the core network. Given the complexity of the scenario presented in Figure 2.1, it is evident that no single MM mechanism will form the solution to all the possible situations and scenarios that will be prevalent. And, although current MM mechanisms propose methods for careful RAT and BS selection, IP packet forwarding, route optimization, and session management, a more than 10-fold increase in user density coupled with the heterogeneity in flow types and network will extremely limit their capabilities. New user applications such as Augmented Reality, Virtual Reality, Vehicle-to-Everything (V2X), etc., will present very restrictive delay requirements, exceptionally high reliability and bandwidth requirements [38], that will consequently severely challenge the capabilities of current MM strategies. Further, the RAN technologies themselves are expected to undergo important transformation in the future networks given the significant interest in VLC, Li-Fi, etc., [32, 33]. Whilst both Li-Fi and VLC, being TeraHertz (THz) bandwidth technologies, enable near Terabits per second (Tbps) speeds, they are significantly impaired by the environment. This consequently has significantly more detrimental effects on the user QoS during mobility scenarios, which we will discuss in further CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 21 critical to the future MM suite. Given the requirements in Table 2.1 and the aforesaid design considerations, a complete overhaul of MM mechanisms for future wireless networks might result in optimal MM solutions. However, the time to develop and market them will be correspondingly longer. Hence, it is prudent to explore and evaluate the myriad legacy as well as current state-of-the-art mechanisms and standardization efforts, and evaluate their suitability as enablers for MM in 5G and beyond wireless networks. But, before we perform such an analysis, a detailed background into these mechanisms has been provided in Sections 2.2 and 2.3. Concretely, we have divided the existing mechanisms into two categories, i.e., Legacy mechanisms (2G/3G/4G defined by 3GPP and also non-3GPP solutions such as Wi-Fi) and Current mechanisms (5G as defined by 3GPP and other relevant solutions proposed by academic and standards bodies). 2.2 Mobility Management: Legacy Mechanisms Mobility Management, as has already been stated, permits a user to stay connected even when it moves beyond the geographic boundaries of the network to where it first attached to. This property, as a result, also determines the ubiquity of a given wireless standard. Further, given the softwarized characteristic of the 5G and beyond networks, and the centralized nature of current and legacy mobility management strategies, it becomes even more critical to explore the various avenues/aspects of future MM strategies. Henceforth, as part of our detailed study, we first reflect back on some of the most significant MM strategies that have served the wireless networks well up until now. Concretely, in the following subsections we present a discussion on the MM strategies for pre-5G networks. 2.2.1 3GPP based MM techniques In this subsection, the various mobility management techniques developed and deployed by 3GPP, as part of the LTE framework, have been discussed in detail. 2.2.1.1 LTE Handover Mechanisms 3GPP based LTE [59] standard is the most widely accepted and subscribed wireless standard today. With the global subscription reaching 1,100 million [60] for LTE in 2015 and expected to increase by four times by 2021, it is imperative to understand the mobility management mechanisms as prescribed by 3GPP for LTE. CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 22 eNB SAE-GW MME PCRF Internet Figure 2.2: Basic LTE Architecture. In order to understand mobility management in LTE, it is essential to first understand the LTE core structure and the entities that perform mobility management. Figure 2.2 provides a diagrammatic representation of the LTE architecture. From Figure 2.2, the Mobility Management Entity (MME), as the name suggests, is responsible for the handover and policy management functions. It not only performs the negotiations for resources in the core and access networks (in case of S1 handover) but it is also responsible for communicating with the Policy and Charging Rules Function (PCRF) and the System Architecture EvolutionGateway (SAE-GW). Note that, the SAE-GW is a single entity representing both Serving Gateway (S-GW) and Packet Gateway (P-GW) together as one unit. And so, in addition to the MME, P-GW is another entity that is connected with the mobility management of the UE. It is known that the LTE network is an all-IP network, and hence, every UE that accesses the network receives an IP address. This IP address is assigned by the P-GW and it acts as the anchor for the same until the UE stays in its domain. Note that, layer 3 mobility is an important component in ensuring continuity of the service whilst allowing mobility. This is so because, many application services are not designed to handle a change in IP address without service interruption. And hence, techniques to allow for seamless mobility in such scenarios is of great interest when studying mobility management. It is important to state here that mobility management in LTE involves: 1) Handover (when the UE is in active state) and 2) Cell re-selection and Tracking Area Update (TAU) (when the UE is in idle state). Handovers: In LTE, handover may either be within the same Tracking Area (TA) (or CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 23 a TA registered in the MME) or it might be to a TA that is not associated with the serving MME, and hence may entail the extra step of tracking area registration. Further, LTE provides two types of handovers, i.e., X2 HO and S1 HO [61]. Whilst within the same tracking area and performing a HO, the UE first sends its measurement updates to the Source eNB (SeNB). The SeNB is the eNB to which the UE is currently attached to. And so, if the measurement report results in the decision to HO, the SeNB then checks for the presence of the X2 interface to the target eNB (TeNB). TeNB is the eNB that is chosen to which the UE has to be handed over. Thus, in case it is absent, S1 HO is initiated. For S1 HO (illustrated in Figure 2.3), the SeNB informs the MME about its decision, which in turn informs the TeNB about the HO request. After negotiating for the resources, the TeNB formulates an indirect route to the SAE-GW and informs the MME about it. An indirect route refers to a route that allows the SeNB to tunnel the packets to the TeNB via the SAE-GW. Hence, the MME informs the SAE-GW about the same, and also asks the SeNB to form an indirect route with the SAE-GW. This allows for the SeNB to tunnel the downlink packets to the TeNB whilst the HO is being executed. Further, the SeNB also informs the TeNB about the sequence number of the downlink and uplink packets. Next, the SeNB signals the UE to perform the HO, and starts tunnelling its DL packets to the TeNB where they are buffered. After the HO is completed, the UE indicates to the TeNB that it has completed the HO. This allows the TeNB to start transmitting the buffered DL packets and accept the UL flow. Further, the TeNB informs the MME about the same, which then informs the SAE-GW and SeNB. The SAE-GW then switches the path to the most optimized one, i.e., it now invalidates the indirect route, and at the same time SeNB releases the UE context. In this way the LTE S1 HO is executed and complete. An illustrative description of the S1 HO process is presented in Figure 2.3. However, and as mentioned above, in case the SeNB has an X2 interface to the TeNB, the HO negotiation is performed through the X2 interface. The benefit of the X2 interface is that it does not entail any signaling with the elements in the Evolved Packet Core (EPC), and hence it reduces the signaling load in the core network as well as reduces the interruption time. Consequently, the QoS for the users is also improved. The drawback to this method is that to have an X2 interface eNBs need to be connected to each other either through fibre or microwave links, which increases the CAPEX and OPEX for the service provider. When X2 HO is the selected method, SeNB passes the HO request message to the TeNB. After resource negotiation, the TeNB sends back a HO acknowledgement message to the SeNB. Further, the SeNB now informs the UE to start the HO, and tunnels the DL packets to the TeNB. As soon as the HO is confirmed, and TeNB receives this message from the UE, it informs the MME of the event. The MME now requests the SAE-GW to switch the CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 24 S1 HO required SeNB Link measurement HO request TeNB MME SAE-GW UE HO Command SeNB TeNB MME SAE-GW UE Previous Data Path To CN DL Buffering SeNB HO Complete TeNB MME SAE-GW UE HO Complete Resource Release Path Switch New Data Path Figure 2.3: 3GPP S1 Handover CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 25 paths, and after the path has been switched for the DL, a release resource message is sent to the SeNB. This causes the SeNB to release all the resources reserved for the UE that just completed HO from its domain to that of the TeNB. And hence, in this way the LTE X2 HO is executed. Similar to Figure 2.3, in Figure 2.4 an illustrative description for the X2 HO is provided. Cell re-selection and Tracking area update: Cell re-selection and TAU happen when the UE is idle, i.e., it does not have any active sessions running. The process of cell reselection is fairly simple process (here we initially assume that the UE moves within the TAs that are already registered at the MME). Firstly, the UE while camping on a particular cell performs neighbour cell and serving cell measurements. After ranking the cells based on the RSSI, if the serving cell RSSI is greater than the threshold then no re-selection is performed. However, in the case when the neighbouring cell RSSI is greater than the threshold as well as that of the serving cell, cell re-selection is performed and the UE now camps on the new cell. Further, in the event that this new cell is in a TA that is different from those registered at the MME, a TAU message is sent to the MME. The MME then registers the new TA and also sends back a list of TAs to the UE. This list allows the UE to traverse in the TAs mentioned without performing any TA update. In the event, that the UE is not idle and it performs a HO, a HO with TAU is performed. In this case, the HO procedure is the same as mentioned before, the only additional signaling being that at the end of the HO a TAU, as discussed above, is performed. 2.2.1.2 3GPP Dual Connectivity, LTE-WLAN Aggregation and LWIP The Dual Connectivity (DC) concept allows a user to camp on two BSs simultaneously. Concretely, a UE can be connected to a Small-cell (SC) and a Macro-cell (MC) at the same time, wherein the MC and SC are connected to each other via the X2 interface. According to 3GPP, all control plane communications, including resource allocation on SC, are performed via the corresponding MC, to which the UE is associated to. Note that, DC was introduced by 3GPP for LTE during Release-12. But, it is in Release-13 that this concept matured, wherein multiple usage scenarios, architecture and the operational characteristics were defined. A detailed description of the same has been presented in [62]. Furthermore, during Release-13, the concept of LTE-WLAN aggregation (LWA) was standardized [15]. Through LWA, a UE can simultaneously receive packets over both the LTE and the Wi-Fi interfaces, wherein the aggregation (and splitting in the eNB) of these two physically distinct data streams takes place at the Packet Data Convergence Protocol (PDCP) layer in the protocol stack. However, note that the LWA functionality is defined only for the downlink [63]. CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 26 HO Command SeNB SAE-GW SeNB Link measurement HO request TeNB MME UE X2 Interface SeNB TeNB MME SAE-GW UE Previous Data Path To CN DL Buffering HO Complete TeNB MME SAE-GW UE HO Complete Resource Release Path Switch New Data Path Figure 2.4: 3GPP X2 Handover CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 27 Another technique, similar to LWA, is the LTE WLAN integration using IP security tunnel (LWIP) [64]. In this technique, while the objective is similar to that of LWA, i.e., to integrate the LTE and WLAN technologies, it is done at the network layer in LWIP. However in LWA, the integration is performed at the PDCP layer. Further, LWIP, unlike LWA, can be implemented for both uplink and downlink. 2.2.1.3 3GPP Traffic Offloading Traffic offloading essentially helps the operator to reduce the amount of traffic in its core network [65] by either offloading it to another domain in its network or to a completely different network such as Wi-Fi, the latter of which will be explored in more detail later. And so, specifically with this objective 3GPP introduced the Local IP Access (LIPA) and Selected IP Traffic Offload (SIPTO) frameworks [53, 65, 66]. The main reason to analyze the traffic offload frameworks is that they implicitly involve mobility from one domain to the other. And hence, the requirements of traffic offloading also need to be taken into consideration whilst designing the mobility management scheme and policies. 3GPP through the LIPA and SIPTO framework made provisions for traffic offloading approach in 3GPP networks. LIPA allows the network to offload the traffic locally if both the mobile node (MN) and the Correspondent Node (CrN) are in the same domain. By same domain we refer to the fact that the MN and CrN are associated with the same Home eNB (HeNB). Figure 2.5 presents the scenario involving LIPA traffic offload. Additionally, LIPA also allows for a local breakout to the Internet/CrN domain, hence bypassing the core network during the flow of data. An important challenge of LIPA with regards to MM is that, session continuity for LIPA connections during mobility events is not supported. Further, the SIPTO framework enables the network to offload traffic to a geographically proximal gateway, hence, allowing the network to reduce traffic load on a particular gateway. Figure 2.6 provides an illustrative description as to how SIPTO framework operates. As can be seen from Figure 2.6, the traffic to the UE is offloaded to a set of gateways that are geographically close to the UE’s point of attachment to the access network, which here are P-GW2 and S-GW [53]. Next, during 3GPP Release-10, the concept of IP Flow Mobility (IFOM) was also introduced. IFOM allows a UE to offload, if possible, data sessions to the Wi-Fi network from the 3GPP network. Consequently, through IFOM, a UE can maintain data flows belonging to the same packet data network (PDN) connection simultaneously on both the 3GPP and the Wi-Fi network [66]. However, while in LWIP the connection eventually passes through the LTE core network, in IFOM the offloaded connections pass through the WLAN network, and onto the IMS core. CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 28 UE UE HeNB S-GW P-GW Internet Local breakout Local traffic Figure 2.5: Local IP Access (LIPA) SIPTO Traffic P-GW2 P-GW1 S-GW eNB UE CN Traffic Figure 2.6: Selected IP Traffic Offload It is important to re-iterate that since traffic offloading procedures implicitly invoke mobility management schemes/policies, it becomes necessary to study them with the perspective of designing mobility management schemes of 5G and beyond networks, which will be done in Chapter 3. CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 29 2.2.2 ITU – Vertical multi-homing The future generation of wireless networks is envisioned to be one that is both dense in users as well as BSs, and heterogeneous. By heterogeneity it is understood that the network will comprise of multiple RATs co-existing in a single domain. This provides an opportunity to the users to utilize multiple RATs in order to improve the total throughput, reduce latency and increase the reliability of their link. And so, the ITU-T through its study on vertical multi-homing [67] provides the requirements, expectations and an architecture to perform the same. Concretely, in vertical multi-homing, each layer has a multi-homing feature and there are many network resources used to establish multiple network connections. In PHY/MAC layer, multiple network access technologies, multiple network interfaces, multiple channels, and multiple radios are network resources. In network layer, multiple IPv6 addresses and multiple prefix information are network resources. In transport layer, multiple transport sessions are network resources. To efficiently establish multiple network connections and manage network resources, it is needed to manage them in an integrated and harmonized fashion. Moreover, while vertical multi-homing consists of aspects on how to connect with multiple interfaces (and how support across layers for the same is provided), MM in multi-RAT and multi-connectivity scenarios becomes a complex issue and hence, an analysis into the vertical multi-homing concept is well placed in the realm of our current research. To elaborate, according to [67], vertical multi-homing entails having multi-homing capabilities in each of the layers of the implemented OSI network model, at both the host and the client. The modular approach of the OSI model, although ideal for making modifications without affecting other blocks, is the first challenge that vertical multi-homing encounters. This is so because, inter-layer coordination to optimize resource allocation and utilization is essential for the purpose of vertical multi-homing. As an example, consider the PHY/MAC layer has multiple interfaces active (each interface corresponds to a different technology) and the network layer has multiple active IP prefixes, however if the layers do not interact with each other, then the network layer will not know about the multiple interfaces. Consequently, an optimal association between active IP prefixes and active network interfaces cannot be determined, which will lead to an inefficient utilization of the resources. Hence, it is imperative that the protocol layers interact with each other. Further, as additional requirements, vertical multi-homing necessitates the provision of routing optimization, QoS based connection selection in presence of multiple network connections, bandwidth utilization over multiple network connections (through optimal stream splitting and combining), recovery methods in the event of network interface failure, and network interface selection (in the CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 30 event a single or a small subset of available networks can be utilized). Additionally, [67] also presents some methods that can be employed to implement multi-homing. These methods are classified as: 1) Based on correspondence between IPv6 address and interfaces (multiple IP addresses may configure multiple interfaces or a single IP address may be shared amongst multiple interfaces), and 2) Based on the supporting layers for multi-homing. Through the requirements and methodologies as mentioned above, it is evident that the vertical multi-homing implicitly invokes mobility management schemes/policies. Moreover, as a step to handle vertical multi-homing on the network/host side a vertical multi-homing functions block, represented in Figure 2.7, is defined by ITU. Resource Identifying function Resource management function Network status recognize and adjust function Interaction across layers function Figure 2.7: ITU-VMH functions block Concretely, the vertical multi-homing functions such as resource identifying function, resource management function, network status recognize and adjust function, and lastly interaction across layer function provide functionality not only for vertical multi-homing, but they also serve as resources for carrying out mobility management whilst enforcing vertical multi-homing. And hence, the ITU-T, through its study in reference [67], implicitly tackles certain mobility management aspects such as RAT selection, resource identification and management, network status feedback and adaptive management, to mention a few. 2.2.3 CoMP The Co-ordinated Multipoint (CoMP) strategy involves multiple base stations co-ordinating with each other to serve a given user [68]. Similar to ITU-VMH, CoMP can provision CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 37 legacy systems, proxy servers supporting MPTCP will need to be installed in front of the legacy devices, such as the middleboxes installed by service providers. The legacy systems can then communicate with the proxies using the legacy TCP protocol, while the proxies utilize MPTCP for communicating with the destination MPTCP capable device. Such a requirement will potentially impact the scalability of the MPTCP solution for 5G and beyond MM mechanisms. 2.2.4.6 SCTP Stream Control Transmission Protocol (SCTP) [80], like MPTCP supports multi-homing and allows for separate message streams to be sent simultaneously. However, it differs from MPTCP in the fact that it incorporates qualities of both User Datagram Protocol (UDP) and Transmission Control Protocol (TCP) in how the messages are handled, whereas MPTCP is a direct extension of TCP. Additionally, and similar to MPTCP, it provisions multipath redundancy and congestion awareness [90]. Moreover, it also facilitates flow level granularity of service, like MPTCP, which will be important for 5G and beyond MM mechanisms. Hence, SCTP through its features is also a potential future MM mechanism enabler. Note that, for SCTP, both the user and server protocol stacks need to be updated [90]. The aforesaid update will essentially be a software update, wherein the transport layer of the protocol stack is updated. However, given the number of users in future networks, it will pose a scalability challenge for the deployment of SCTP as part of the 5G and beyond MM mechanisms. 2.2.5 IEEE Media Independent Handover 802.21 In order to provide inter-domain mobility, i.e., between various IEEE 802 standards as well as non-IEEE 802 standards such as 3GPP technologies, IEEE standards group came up with the IEEE 802.21 standard. The IEEE 802.21 is a Media Independent Handover (MIH) service, that as the name suggests provides a common intermediate platform and consequently enables the upper layers to interact with the lower layers (layer 2 and below) irrespective of the technology [91]. The MIH service provides the event, command and information services, which form the core of this protocol. This enables the higher layers in the protocol stack to query information that is present on the link and MAC layers for ensuring seamless connectivity in between domains and consequently enhance the user QoE [43,91–94]. Further, the IEEE 802.21c amendment [93], provides insights as to how UEs with single CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 38 radio can perform seamless inter-domain handovers. The suggested approach is a makebefore-break approach. This amendment also states that the most time consuming process when involving an inter-domain HO is the authentication and context information exchange. And hence, in order to reduce this latency a proxy connection approach is adopted, depicted in the architecture in Figure 2.12. In Figure 2.12, the MN, which is undergoing a handover, is initially attached to the source network via a Source Point of Attachment (Source PoA). The MN is consequently being handed over to the Target PoA (TPoA). And so, the MN, instead of traversing the entire network to reach the information server, to retrieve details regarding the candidate target networks and their corresponding handover policies, accesses the same via the Source Point of service (SPoS). Following the handover decision, it is the responsibility of the SPoA to communicate with the TPoA, with regards to resource allocation, and authentication and context information exchange. These processes, according to [93], are performed proactively via the interaction between the SPoS/SPoA and TPoS/Proxy TPoA. Target PoA TPoS/ Proxy PoA Source Network Target Network Information Server MN during handover Source PoA SPoS/Proxy Info Server Figure 2.12: IEEE 802.21c – Single Radio handover functional model Thus, through the provisioned proactive information transfer between the target and source networks, the latency can be significantly reduced. Further, in Figure 2.12, the proxy CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 39 connection between target and source networks helps in the secure exchange of context and UE authentication data, while there is no requirement for a layer 2 attach for exchanging such information. This allows for fast and secure handovers and in this way inter-domain handovers for single radio receivers can be made seamless. It is important to mention that as stated in [95], the initial drafts of IEEE 802.21 did not favour mobile assisted handover mechanisms. However, through the course of its standardization IEEE 802.21 has adopted an approach wherein the decision to perform a handover is taken through collaboration between the network and the UE. And so, with these provisions, IEEE has facilitated the process of providing a standardized platform for the various heterogeneous technologies to co-operate and allow seamless mobility [43, 91–94]. Moreover, 3GPP technologies can also utilize this information and hence, allow devices to handover from 3GPP to IEEE 802.x RATs and vice versa. 2.2.6 RSS based BS selection methods The erstwhile Received Signal Strength (RSS) based methods employ a very simplistic approach to BS selection, by comparing the detected BS link quality (RSSI/RSRP/RSRQ) levels [75,96,97]. The aforesaid simplistic nature renders them easy to implement, and does not entail a high processing and signaling load either. However, such an approach can be plagued by multiple issues. For example, BSs with a good RSS might be overloaded (as more users will be assigned to them) whilst others maybe under-utilized. Such a scenario also implies that a better RSS does not always guarantee better QoS, since, congestion will lead to degraded QoS. Moreover, in dense scenarios, even with the implementation of a hysteresis, UEs will be subject to FHOs due to the fluctuating RSS and availability of multiple candidate BSs. This exemplifies the unreliable nature of RSS based methods for 5G and beyond MM. Additionally, these methods are one-dimensional, given that they consider only RSS as a decision parameter. The RSS methods also do not provision any granularity of service, context awareness, multiple levels of HO support, etc. It is imperative to state here that, in Chapter 3, wherein we present the qualitative gap analysis, a subset of the legacy mechanisms discussed in this section have been analyzed. Concretely, we analyze the 3GPP LTE handover mechanisms, 3GPP LTE Traffic offloading, 3GPP Dual Connectivity and LWA, IETF PMIPv6, IETF MPTCP, IETF SCTP, IEEE 802.21 and RSS based BS selection methods. We choose the aforementioned strategies for the qualitative gap analysis due to their wide-ranging acceptance/applicability in wireless networks. CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 40 2.3 Mobility Management: Current State of the Art Global efforts have spinned up consortiums that have provided impetus to the development of 5G, including that of MM strategies. Further, for B5G networks, such as 6G, certain collaborative efforts have already started. References [32,33,35,36,98] highlight the advances that have been made with regards to identifying the enablers and core principles of B5G networks. Hence, in this section we first detail the 5G architecture defined by 3GPP, followed by the discussion on current state of the art in MM mechanisms. 2.3.1 3GPP 5G Architecture Background We introduce, through Figure 2.13, the 5G architecture standardized by 3GPP [45]. Concurrently, we have also presented the classification of the various mechanisms that we explore in Sections 2.3.2 and 2.3.3 with respect to the 5G architecture in Figure 2.13. This classification is dependent on the portion of the network that is impacted (directly or indirectly) the most by a particular MM scheme. Furthermore, we have illustrated whether the studied mechanisms are either control plane or data plane solutions. Concretely, a CP solution would primarily impact MM via either CP signaling or decisions, while a DP solution would entail provisioning alternate and more efficient data paths. A detailed discussion with regards to these classifications has been provided in Sections 2.3.2 and 2.3.3. Concretely, the 5G architecture, as shown in Figure 2.13, consists of two main core network functions, i.e., the Session Management Function (SMF) and the Access and Mobility Management Function (AMF). The SMF communicates with the User Plane Function (UPF) over the N4 interface, while the AMF is responsible for communicating with the RAN side over the N2 interface. Furthermore, the AMF and SMF communicate with other network functions, such as the Policy Control Function (PCF), Authentication Server Function (AUSF), etc., to execute their defined functionalities within the ambit of the policies and existing user and network context. For the sake of conciseness, in Figure 2.13 we club all of these functions into a single entity box called Network Functions. Moreover, the AMF also has an N26 interface that connects to the EPC to facilitate Inter-RAT mobility, while an N32 interface exists in the event of a change in Public Land Mobile Network (PLMN) with 5G Core (5GC) as the CN for both the visited and home networks. Note that, the interfaces N2,N4,N26 and N32 are all control plane paths, with the AMF, SMF and other network functions forming the control plane entities. In addition, the AMF in 5G networks is the equivalent of the MME in LTE-4G networks. It focuses on handling mobility at the access network level (such as BS selection, resource allocation, etc.). The SMF on the other hand handles the CN related tasks during mobility CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 41 N4 N6 N2 N2 N2 N4 N4 Other 5GC Network Functions To other 5GC N32 SMF AMF RAN RAN RAN N3 N3 N3 UPF UPF UPF N9 N9 UPF N4 IMS Core Local breakout to Internet SDN Based [C] Other Core Network based solutions Edge Clouds [D]  Centralized  Semi-Centralized  Hierarchical  Data Caching based  CP Processing based DMM Based [C/D]  Fully distributed  Partial DMM  SDN based DMM Other Access Network based solutions Cross layer [C] RAN-as-aService [C] Intelligent RAT selection [C/D] Other Extreme Edge Network based solutions Device-to-Device [C] 3GPP 5G MM Solutions [C/D] N26 To EPC C D C/D Data Plane Control Plane Control Plane Solution Data Plane Solution Control + Data Plane Solution Phantom Cell Method [C/D] Core Network Access Network Extreme Edge Network Figure 2.13: Classification of the state of the art in MM strategies on the 5G architecture. CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 42 events (such as path re-routing, etc.). Next, in Figure 2.13, it can be seen that the RAN interacts with the UPF through interface N3, and the UPFs use the N9 interface to communicate amongst themselves. Also, the 5G networks provision a local breakout through the N6 interface from an UPF. The interfaces N3,N9 and N6 constitute the data plane paths, with the RAN and UPF forming the data plane entities. Lastly, the UE, which is also a data plane entity, interacts with the AMF through the N1 interface. However, to maintain clarity, we have omitted the illustration of this interface from Figure 2.13. Thus, with this background, we now explore the current state of the art in MM mechanisms. 2.3.2 3GPP 5G MM Mechanisms 3GPP, through TS 23.501 [45], TS 23.502 [99] and TS 38.300 [100], has provided significant insights into the design and development of 5G MM strategies. New session management methods, service continuity states, UE mobility monitoring, provisioning for multi-homing, load balancing strategies, provision of on-demand MM, resource allocation due to mobility events, the new MM module, i.e., AMF, interand intranext generation core (NGC) handovers, and LTE-EPC 5G-NGC interworking have been introduced in the aforesaid 3GPP specifications. These techniques through the provision of a softwarized solution and a global view of the network scenario alongside user context appear to facilitate the efficient operation of 5G and beyond MM mechanisms. Consequently, in the text that follows, we discuss these newly defined 3GPP MM mechanisms. A. UE Mobility monitoring: In TS 23.501 [45], details with regards to how the UE mobility is monitored and the corresponding actions with regards to resource allocation and context updates have been specified. Concretely, when a UE is mobile, the 5G standards define that the AMF will be responsible for monitoring its movement and hence, its mobility pattern. Furthermore, during a UE mobility event, new resources on the destination BS are managed by the AMF through the RAT and Frequency Selection Parameter (RFSP). Such a process simplifies the identification of the required resources, as well as migration of these resources to the destination network. Moreover, the AMF manages the UE mobility event notification, i.e., it provisions details with regards to the mobility event as well as the areas of interest (Tracking areas, Cells, RANs, etc., to which a UE might migrate to). The other Network Functions (NF), such as the SMF, can subscribe to these notifications so as to employ their decisions and policies. B. Session Management: Through TS 23.501 [45], the various modes that can be utilized CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 43 to manage the multiple heterogeneous sessions for a given user has been defined. Notably, if a UE is connected to multiple RATs then, for a given Protocol Data Unit (PDU) session, the UE has the choice to select the access network over which this PDU session will be served. In addition, the UE, in the event of mobility or congestion, can request a PDU session to be transferred from 3GPP to non-3GPP RAT(s). Furthermore, in roaming scenarios, PDU sessions can either avail a local breakout or be routed through the home network. Specifically, each PDU session can be granted, independently, different routing modes. To do so, the SMF in the 5G CN controls and monitors the status of the data paths. Moreover, the SMF also provisions the capability of performing selective traffic routing by the application of Uplink Classifier (UL CL) on certain data plane entities, i.e., UPFs. A UPF essentially performs the function of a router in the 5G network. C. IPv6 multihoming: The new 5G standards, as specified in TS 23.501 [45], have formalized the use of IPv6 multi-homing so as to reap the benefits from the multiple physical channels that will be available for use through multi-connectivity. Specifically, according to TS 23.501, more than one session anchor can be specified for a PDU session. Note that, a PDU session anchor’s primary role is to assign the IPv6 prefixes that are used by the UE for a given PDU session to communicate with the public network. However, all these PDU session anchors will have a single UPF as a branching point. Next, during a mobility event, a make-before-break approach for a PDU session is adopted to provision service continuity. It must be stated here that, service continuity is ensured through the Session and Service Continuity (SSC) modes, which we will discuss next. D. Session and Service Continuity Modes: 3GPP, through TS 23.501 defines the SSC modes, which are critical for the network in determining the level of service continuity offered to a PDU session [45]. Concretely, three modes are defined: SSC mode 1,SSC mode 2 and SSC mode 3. We briefly describe them as follows: –SSC mode 1: This mode ensures that the IP address is preserved. Specifically, the PDU session anchor is maintained regardless of the access technology being used by the PDU session after the mobility event. Furthermore, the IP address is maintained throughout the lifetime of the PDU session. Additionally, more PDU session anchors might be allocated for additional IP addresses, however, it is not necessary that they be maintained just like the initial IP address and PDU session anchor. CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 44 –SSC mode 2: In this mode, if needed, the network can release a PDU session and request the UE to immediately establish a new PDU session with the same network. Moreover, if the UE has multiple PDU session anchors, the additional anchors can be released or allocated (for new IP prefixes/addresses). –SSC mode 3: In this mode, IP address is not preserved. This consequently makes any changes in the user plane visible to the UE. However, to ensure that an acceptable level of QoS, and hence, service continuity is maintained, a makebefore-break approach is followed. This essentially determines the destination PDU session anchor before relieving the resources the PDU session occupies at its current anchor. It must be stated here that the SSC mode for a UE is selected by the SMF depending on the UE subscription details as well as the PDU session type. E. User Plane aspects: In 5G networks, UPFs will be utilized to handle the data plane traffic. Concretely, they can be thought of as routers, on whom the routing rules are programmed by the SMF. In TS 23.501 [45], the aforesaid specifics have been defined. However, note that the methodology to establish these paths still involves exchanging Tunnel Endpoint Identifiers (TEIDs) between CN entities. This, as we will state in the analysis, can be a cause of increased network load. Additionally, traffic re-routing, in the event of mobility or load balancing, is handled by the SMF, wherein it sends the necessary information, such as the forwarding target information, to the UPFs. Lastly, in the event of mobility of a UE, packet buffering is also provisioned so as to minimize the loss of packets and hence, QoS. F. Dual Connectivity: Through TS 23.501 [45] and TS 37.340 [101], 3GPP has also concretized and standardized the integration of Multi-RAT Dual Connectivity (MR-DC) into 5G. Concretely, the UEs will now have the capability and possibility to connect to two BSs belonging to the same RAT (LTE-LTE, 5G New Radio (NR) - 5G NR) or to different RAT(s) (LTE - 5G NR). As in LTE-DC, this can be configured to allow fast-switching (fast HO), since control plane is not changed unless the Master Node is changed. Further, it can also be used to increase data rates by using both RATs at the same time. Also, the UP is terminated at Master node, so, no CN signalling is necessary for intra-Master Node HO. G. Edge Computing: TS 23.501 [45] defines the support for edge computing platforms in 5G networks. Concretely, these are utilized in the non-roaming or local breakout CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 45 roaming modes. Note that, the 5G CN is responsible for selecting a UPF that is close to the UE and also has access to an edge compute node. Consequently, traffic steering is performed at this UPF towards the edge compute node. H. Network Slicing: The concept of enabling a telecom operator to be able to slice its infrastructure network into logically separated networks and consequently service multiple tenants, e.g., virtual network operators, services (eMBB, URLLC, mMTC), etc., using the same, wherein the logical separation involves dynamic allocation of network resources, is termed as network slicing [57]. 3GPP, in TS 23.501 [45], has discussed network slicing in detail, wherein its support for roaming as well as its involvement in the inter-working process between 5G CN and LTE-EPC have been elaborated. It also defines support for migrating and translating the Single Network Slice Selection Assistance Information (S-NSSAI), which consists of the necessary information with regards to an assigned network slice for a UE, between the Home PLMN (H-PLMN) and the Visited PLMN (V-PLMN) has been detailed. Similarly, for the inter-working process, 3GPP charts out the principles for migration, translation and creation of SNSSAIs whenever a UE undergoes mobility and changes from a 5G network to an LTE network, and vice versa. Moreover, the support has been defined for scenarios where the N26 interface, which is the standard 5G CN and LTE-EPC inter-working interface, may or may not be present [45]. On the other hand, and importantly, the concept of network slicing also assists in provisioning tailor-made MM solutions for the tenants that each network slice will cater to. This consequently helps to deploy on-demand MM strategies. I. Load Balancing and Congestion Awareness: In TS 23.501 [45], 3GPP has defined procedures for load balancing at the AMF and SMF, as well as congestion awareness within the core network. Concretely, two specific strategies, i.e., load balancing and load re-balancing, have been provisioned. Within the load balancing paradigm, new users incoming into an AMF region, if necessary, are directed to an appropriate AMF in order to manage the load of the AMFs. To do this, appropriate weights, indicative of the load on each AMF, are assigned and updated at appropriate intervals (typically on a monthly basis). On the other hand, if an AMF becomes overloaded, then load rebalancing is performed. Here, already registered users are migrated to other AMFs that are not overloaded while ensuring minimum service disruption [45]. Note that, the new AMF chosen should belong to the same AMF set. An AMF set is defined as the AMFs which belong to the same PLMN, have the same AMF region ID and the same AMF set ID value [45]. These parameters are pre-configured by the network operator. Lastly, CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 46 3GPP also provisions extensive details with regards to handling congestion control for the Non Access Stratum (NAS) messages. This is important from the perspective of MM, as MM messages are carried over NAS to the CN nodes. For further details with regards to the specifics of the congestion control procedures, the reader is referred to TS 23.501 [45]. J. Cell, Beam and Network Selection: Through TS 23.501 [45] and in particular through TS 38.300 [100] details with regards to cell, beam and network selection have been specified. For cell selection these standards documents, developed by 3GPP, specify support for cell selection procedures given that the UE is in either Radio Resource Control (RRC) idle, or RRC inactive or RRC connected state. Note that, RRC idle state refers to a UE that can listen to paging channels, broadcasts and multicasts, as well as perform cell quality measurements. The RRC inactive state refers to a UE that, in addition to the functionalities specified in the RRC idle state, can roam within the RAN-based notification area (RNA) without informing the NG-RAN. The RRC connected state for a UE implies that it has an active connection and data flow. Most notably, for the RRC connected state, cell mobility and beam mobility have been specified. As the name suggests, a UE can either undergo a cell handover or it can switch between the beams that a given BS uses. To perform this, procedures for beam quality and cell quality measurements have also been defined in [100]. The beam quality measurements are performed in the physical layer for multiple beams being transmitted by a given cell. These measurements are filtered and aggregated at the RRC layer to obtain the cell quality measurements. Note that, these quality measurements are still performed using the RSSI/RSRP/RSRQ/SINR metrics. Furthermore, in [100], procedures for cell selection and handover involving intraand inter-frequency handover in 5G NR, Inter-RAT handover within 5G CN, Inter-RAT handover from 5GC to EPC and vice versa, have been specified. We refer the reader to TS 38.300 [100] for a more detailed discussion on the same. Moreover for Inter-RAT handovers, procedures for packet buffering and forwarding as well as data path switching, to ensure the requested QoS, have also been defined. Lastly, roaming and access restrictions are also appropriately defined based on the user subscription to both the SMF and AMF. This facilitates the selection of the right BS and PLMN for a given user [45,100]. K. Inter-Working, Migration and Handover signaling: While TS 38.300 [100] specified certain handover procedures for both the CP and DP, a detailed description of the handover signaling, inter-working between 5G CN and EPC, and migration of PDU sessions has been provided in TS 23.502 [99] and TS 23.501 [45]. Concretely, through CHAPTER 2. STATE OF THE ART IN MOBILITY MANAGEMENT 53 required with regards to the future MM solutions. We then detailed the various legacy mechanisms, wherein we explored the strategies developed/standardized by the various standardization bodies like 3GPP, IETF and IEEE. We also, studied strategies developed by the academic community for the same. Following this, we then explored the current state of the art in MM strategies, wherein we presented a detailed discussion with regards to the 3GPP 5G MM mechanisms as well as other research efforts in industry and academia. Concurrently, we also presented a novel classification of these studies based on them being either Core Network based,Access Network Based or Extreme Edge Network Based. Notably, our discussions in this chapter have also taken cognizance of the fact that there have already been some meaningful studies towards Beyond 5G networks and their enablers. Next, we utilize the background developed in this chapter to present a novel Qualitative Gap Analysis for some of the well known/utilized MM strategies in Chapter 3. Chapter 3 Qualitative Gap Analysis in Mobility Management Overview In this chapter, we firstly elaborate upon the three pillars of any future MM strategy, i.e., reliability, flexibility and scalability criteria. We establish a novel relationship between the requirements defined in Chapter 2 and the aforesaid criteria. We then present a novel discussion on the readiness of MM for 5G and B5G networks. We perform this through a novel qualitative gap analysis, wherein we evaluate the pros and cons of the legacy and current MM mechanisms and determine the extent to which they satisfy the aforesaid criteria. Note that, and as we will show in this chapter, a complete agreement towards these three criteria will be equivalent to satisfying the requirements enlisted in Table 2.1. Subsequently, we then determine the persistent challenges that exist towards the development of 5G and beyond MM strategies as well as the potential solutions that will assist in tackling these challenges. We then provision a future framework for the 5G and beyond MM mechanisms. Lastly, we highlight how our contributions, detailed in the Chapters 4-6, aim to realize this framework. Contributions [J2] A. Jain, E. Lopez-Aguilera, and I. Demirkol, "Are Mobility Management Solutions Ready for 5G and Beyond?", Accepted in Elsevier Computer Communications, pp. 1–36, 2020. (Quartile: Q2; IF: 2.816 (2019)) In Chapter 2, we established a detailed background on mobility management mechanisms and their corresponding utility/impact. Following this, it becomes prudent that we analyze 54 CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 55 the gaps that still exist with regards to the MM strategies that will satisfy the 5G and beyond networks’ requirements. And so, in this chapter we present a novel qualitative gap analysis, which is also published in part in [J2]. We evaluate certain predominant legacy mechanisms as well as the current state-of-the-art mechanisms on the basis of reliability, flexibility and scalability: the three pillars of any future MM strategy. 3.1 Qualitative Analysis Criteria As part of this qualitative analysis, we firstly present a detailed description of the three criteria, i.e., reliability, flexibility and scalability, as follows: •Reliability helps to determine whether the MM mechanisms employed will be able to ensure guaranteed and continuous service in any given network topology. Such reliability requirements entail not only continuous connectivity whilst traversing a geographic area, they also include reliability in delivery of packets for critical and delay sensitive services. Further, reliability from a MM mechanism also envelops factors such as tolerance to congestion (through for example, Distributed MM), ensuring faster yet trustworthy re-connection and authentication whilst mobile, ensuring appropriate levels of redundancy in the number of flows, connections, and hosts, and also ensuring appropriate resource allocation for users with myriad mobility and application profiles at the edge, access and core network. •Flexibility as a qualitative analysis metric helps to determine the adaptability that MM mechanisms will provide to the network, which as discussed will be heterogeneous and dense in all perceivable aspects. The flexibility provisioned by MM mechanisms for future networks hence envelops factors such as the ability to formulate and deploy MM policies depending on individual user profiles, flow profiles or based on a slice profile. Further, ensuring the possibility of multi-connectivity through various layers, such as transport layer (SCTP/MPTCP), IP layer (Multi-homing), MAC-PHY layer (Dual Connectivity), will be an important factor for ensuring a flexible MM policy. Additionally, factors such as multi-objective base station selection/user association taking into account factors such as congestion, QoS requirements, backhaul reliability, etc., will be critical to a flexible MM mechanism. •Scalability aspect allows one to determine if the future MM mechanisms can serve the increasing number of user devices with a corresponding increase in requested QoS with heterogeneous mobility profiles. A measure of scalability of MM mechanisms can be CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 56 gained by analyzing factors such as number of connections that can be managed given an increasing number of user devices, management of the signaling load generated due to mobility events, management of the increasing load due to processing the many CP messages generated in mobility events, as well as the ability to permit de-centralization (which in essence would ensure scalability) and being easily deployable on a large scale given a new MM mechanism. We summarize the aforesaid criteria into a list of parameters for each criteria and present them in Table 3.1. Additionally, we also indicate the requirements (from Table 2.1) for whose fulfilment each of these parameters contribute towards. Note that, compliance with each of the stated parameters in Table 3.1 for the reliability, flexibility and scalability criteria will be essential towards ensuring that the MM mechanism under consideration satisfies the requirements defined for the upcoming 5G and beyond networks (Table 2.1). We now elaborate upon the parameter-requirement relationships that have been illustrated in Table 3.1, with the objective of enhancing the comprehensiveness of the evaluation criteria. 3.1.1 Reliability: Parameter to Requirement mapping The provision of redundancy in the number of flows and connections, i.e., by satisfying parameter RL1, can help fulfil requirement R7 presented in Table 2.1. This is so because, redundancy in connections will help overcome the fragile nature of wireless channels in the frequency bands that constitute VLC and mmWave communications. Next, satisfying the parameter RL2 will contribute towards fulfilling the requirements R1,R7, and R8 (Table 2.1). Here, the ability to provision seamless handover assists in supporting mobility amongst multiple RAT(s) (R1), supporting multi-connectivity and thus reliability (R7), and utilize enhanced localization capabilities to accomplish the same in dense urban scenarios (R8). Additionally, the RL3 parameter for the reliability criteria, when satisfied, will help to fulfill the R3 and R4 requirements (Table 2.1). The reason being, decentralization will allow for efficient handling of the number of devices (R3). Moreover, to establish an effective level of decentralization, such as for accessing cached data at the edge and in the IMS core, enablers such as NFV and Mobile Edge Computing (MEC) will be utilized (R4). Furthermore, the RL4 parameter holds significant relevance towards fulfilling the requirements R5 and R10 (Table 2.1). Specifically, fast path re-routing in the CN ensures that the increased dynamism, due to the mobility of both the UE and BSs (R5), is catered to in the CN. In addition, data path modifications due to service migration and service replications, which do not lead to extensive delays, is also ensured through parameter RL4. Lastly, satisfying the RL5 parameter will help towards fulfilling the R2 requirement (Table 2.1), since guaranteeing CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 57 Table 3.1: Governing Parameters for the Reliability, Scalability and Flexibility of a MM mechanism/standard # Reliability Contr. to Reqs. # Flexibility Contr. to Reqs. # Scalability Contr. to Reqs. RL1. Redundancy in the number of flows, connections, etc. R7 FL1. Granularity of service. E.g. per flow, per connection, per user, etc. R9, R11 SL1. Manageable number of connections with increasing number of users R3, R9 RL2. Seamless handover capability† R1, R7, R8 FL2. Capability to enable connectivity to multiple BSs R1, R9 SL2. Manageable signaling load with increasing number of users R3, R9 RL3. Decentralization R3, R4 FL3. Handover service support at multiple network levels. E.g. Core network, Access network, etc. R4, R9 SL3. Manageable processing load with increasing number of users/devices R3, R9 RL4. Fast path re-routing at CN R5, R10 FL4. Handover decision making utilizing multiple parameters. E.g. network load, requested QoS, etc. R1, R9 SL4. Decentralization R4 RL5. Congestion aware R2 FL5. Context awareness R2, R9, R10 SL5. Ease of implementation and integration R6 †Seamless handover capability refers to the ability of a MM mechanism to permit vertical (inter-RAT) as well as horizontal (intra-RAT) handover. CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 58 congestion awareness helps service the different QoS requirements of the applications, such as virtual reality and emergency services, with better reliability. 3.1.2 Flexibility: Parameter to Requirement mapping When a MM mechanism under study satisfies the flexibility parameter FL1, it correspondingly helps to fulfil the R9 and R11 requirements (Table 2.1). This is so because, FL1 states that a MM mechanism should support granularity of service. This will correspondingly assist in accommodating the multitude of service requirements independently (R9) as well as avoid the one size fits all approach (R11). Next, FL2 parameter will help in satisfying the R1 and R9 requirements (Table 2.1). Essentially, the capability to be able to connect with multiple BSs will assist in multi-RAT MM (R1) as well as in provisioning enhanced agility for MM mechanisms in a dense and heterogeneous network (R9). Further, when the FL3 parameter is satisfied, it helps to fulfil the R4 and R9 requirements. The reason being, to enable handover support at multiple levels of the network, usage of SDN, NFV and MEC platform will be necessitated for efficient implementation (R4). Moreover, such multi-level handover support will also provision flexibility for the network (R9). Additionally, satisfying parameter FL4 enables the MM mechanism under study to contribute towards satisfying the R1 and R9 requirements (Table 2.1). Specifically, having a handover decision mechanism that utilizes multiple parameters aids in handling MM amongst multiple RAT(s) more flexibly and hence, efficiently (R1). Also, such strategies will ensure that alongside being flexible, solutions are computationally tractable and energy efficient (R9). Finally, parameter FL5, when satisfied, will be relevant for the fulfilment of requirements R2,R9 and R10 (Table 2.1). To elaborate, the context awareness feature of a MM mechanism will assist in provisioning MM support dependent on application, user and network context (R2), flexibility to handle the increased heterogeneity in the network (R9), and ensure QoS whilst performing complex tasks such as migrating or relocating services based on user mobility events (R10) through appropriate path and resource management. 3.1.3 Scalability: Parameter to Requirement mapping For the scalability criteria, when parameter SL1,SL2 and SL3 are satisfied by a MM mechanism, they correspondingly also assist in fulfilling the R3 and R9 requirements (Table 2.1). Concretely, the ability to be able to manage increasing number of connections, signaling load and processing load with the number of increasing users will correspondingly assist in handling a user density of more than 106devices per km2in 5G and beyond networks (R3). Also, they will help in ensuring the required scalability to accommodate the increasing CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 59 heterogeneity in the network as well as the corresponding tractability of the MM solution (R9). Next, when parameter SL4 for the scalability criterion is met, it helps to fulfil the R4 requirement (Table 2.1). Specifically, to accomplish decentralization objective the MM mechanism under study will need to utilize enablers such as NFV and MEC. Lastly, satisfying parameter SL5 will help to meet the requirement R6 (Table 2.1). The reason being that, ease of implementation usually arises from the fact that a MM mechanism has been used/deployed before, as well as is suitable to accommodate legacy devices whilst catering to a new set of service and devices. Hence, satisfying the SL5 parameter will assist in ensuring that backwards compatibility requirements (R6) are adhered to. And so, from the aforementioned elaborate understanding of the mapping, it can be deduced that the criteria chosen for our qualitative analysis are comprehensive in nature and approach. Moreover, and considering only the 5G networks since their KPIs have been defined [48], provisioning beyond 99.999% reliability will be ensured through the reliability metric during mobility scenarios. Further, latency less than 5 ms for connected cars and 10 ms for virtual reality and broadband applications, will be guaranteed through the reliability and flexibility metric. Specifically, the reliability metric will help provision congestion awareness, reliable link selection, etc., while flexibility will allow multiple type and number of connections during mobility scenarios. In addition, support for nearly 1 million devices per km2with different application and mobility profiles will be ensured through the scalability criterion. Consequently, this further reinforces the comprehensiveness of the criteria chosen for the qualitative analysis that follows. Before we proceed, we highlight certain specifics with regards to the analysis that follows: •The goal of the following analysis is not to compare the considered standards and mechanisms against each other but rather to highlight the extent of their suitability for 5G and beyond networks. •The mechanisms chosen for the qualitative gap analysis are based on their wide-ranging acceptance/applicability in the wireless networks domain. 3.2 Legacy Mechanisms Utilizing the discussions in Chapter 2 with regards to the legacy mechanisms, i.e., Section 2.2, as well as the MM requirements and evaluation criteria specifics in Section 3.1, we now perform the qualitative analysis for the legacy mechanisms in the text that follows. As part of the analysis, for each of the studied mechanisms we firstly highlight their pros and cons CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 60 towards 5G and beyond MM mechanisms. Subsequently, we translate the insights gained from these pros and cons into a summary of parameters satisfied for the reliability, scalability and flexibility criteria. 3.2.1 IETF MPTCP-SCTP Given our objective of determining the suitability of MPTCP and SCTP for 5G and beyond MM mechanisms, we firstly enlist their pros and cons as follows: •MPTCP Pros –Allows for multiple data flows at the transport layer level [78,79,84], and hence, provisions for resiliency against connection failures, given the multipath feature [82–84] –Provisions congestion awareness, with studies such as [86] proposing specific congestion control methods for MPTCP –Through its ability to divide a connection into multiple sub-flows, MPTCP provisions the capability to handle each flow independently [84,88] •MPTCP Cons –The middleboxes installed by service providers are not optimized to support MPTCP [78,79] –MPTCP requires proxies to allow MPTCP enabled devices to take its full benefits [89] •SCTP Pros –Allows for multiple data flows at the transport layer level [85, 90], and hence, provisions for resiliency against connection failures, given the multipath feature –Provisions congestion awareness, wherein reference [90] establishes the presence of congestion avoidance methods within the SCTP suite –Assists in network level fault tolerance through support for multi-homing [85,90] •SCTP Cons –Requires both host and destination device protocols stacks to be updated with the SCTP protocol [90] CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 61 From the pros and cons of both MPTCP and SCTP, as listed above, it can be concretely stated that the IETF MPTCP-SCTP methods satisfy parameters RL1 (allowing for multiple flows over the network for any given user) and RL5 (provisioning congestion awareness as part of the transport layer characteristic for MM) for the reliability criterion. Further, for flexibility, from our discussion above, it is clear that IETF MPTCP-SCTP only satisfies parameter FL1 (by allowing for multiple flows, flow level granularity can be induced). 3.2.2 IEEE 802.21 For the purpose of analysis, we list the pros and cons of the IEEE 802.21 mechanism towards 5G and beyond MM strategies, as follows: •IEEE 802.21 Pros –Provisions seamless handover capability, as it allows users to switch between multiple RATs [43,91,94] –Provisions the possibility for a UE to connect to multiple BSs [43,92] •IEEE 802.21 Cons –Requires the protocol stacks of both the host and destination devices to be modified, so as to enable the IEEE 802.21 functionality [91,93] And so, given the aforesaid pros and cons with regards to IEEE 802.21, it can be deduced that it satisfies parameter RL2 for reliability (allowing for seamless movement between different RATs) and FL2 for flexibility (allowing for the possibility to connect with multiple RATs) criteria. 3.2.3 IETF PMIPv6 Based on the discussions carried out in Section 2.2.4.4, we now enlist the pros and cons of the PMIPv6 strategy with regards to its utility for 5G and beyond MM mechanisms, as follows: •PMIPv6 Pros –Given that PMIPv6 is adopted by 3GPP and it forms a relatively agnostic setup for an UE towards its mobility signaling, it can thus provision seamless mobility [73–75] CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 62 –Through the DMM based PMIPv6 approach, decentralization can be introduced [77]. Furthermore, other approaches, such as the clustering based approach in [76], can grant enhanced scalability and reliability to the PMIPv6 approach –Given that it has already been adopted by 3GPP for LTE, the available implementational expertise will enhance the ease with which it can be adopted in future networks •PMIPv6 Cons –In its original flavor, PMIPv6 suffers from scalability and reliability issues due to the SPoF formed by the LMA in its architecture [76] –An explicit treatment of PMIPv6 with regards to the parameters for flexibility criterion is missing in [73–77] And so, it can be deduced that the IETF PMIPv6 in its original flavor, given its maturity in development and deployment, satisfies the seamless handover parameter RL2 in the reliability criteria. Moreover, with enhancements from the use of DMM and cluster based methods, PMIPv6 can be decentralized and scaled thus satisfying parameters RL3 and SL4 in reliability and scalability, respectively. Furthermore, since it has already been explored and implemented in the LTE networks, it satisfies parameter SL5 owing to its relative ease of implementation as against any other new protocol. 3.2.4 3GPP LTE MM Mechanisms For the 3GPP based MM mechanisms, we firstly highlight the pros and cons for the handover, traffic offloading and DC and LWA strategies, as follows: •LTE Handover Pros –The LTE-X2 and S1 mechanisms together offer handover support at the access and core network level [133] –Through LTE-X2 handover mechanism, CN signaling can be avoided [133] –LTE-X2 permits decision making for a handover to be taken at the access network level. Hence, it reduces the processing load on the CN entities as well and also permits fast handover capabilities [133,134] •LTE Handover Cons CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 69 3.3.2 Other Research Efforts: Core, Access and Extreme Edge Network Solutions 3.3.2.1 Core Network Solutions For analyzing the core network solutions we utilize the generic classifications, i.e., SDN based, DMM based and Edge Cloud solutions, and firstly list their pros and cons. •SDN based mechanism Pros –Provisions global view of the network [106,108] –Provisions hierarchical solutions, thus enabling decentralization [106] –Provisions the ability to manage CN signaling, and hence, DP paths during mobility events [106–108] –Provisions a single point of collection for network statistics thus enabling the design and development of context based MM mechanisms [109] •SDN based mechanism Cons –Extensive CN signaling for managing handovers in a centralized/semi-centralized approach [106] –It does not alleviate the issue of mobility anchors which can lead to SPoFs in the DP •DMM based mechanism Pros –Provisions decentralization of the mobility management anchors [22–24,110] –Assists the CN in implementing efficient data paths for UEs undergoing mobility [22,24,53] •DMM based mechanism Cons –Fully decentralized solution introduces extensive CN signaling in order to manage the changes in data paths and mobility anchors, and hence, handovers [22] –Partially distributed solution, while solving the extensive CN signaling, introduces a central controller, and hence, an SPoF [22] –Co-existence and integration with already deployed networks and devices will be a significant challenge [53] CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 70 •Edge clouds Pros –Ensure data offloading opportunities, and hence, reduction in CN traffic load [25,115] –Facilitate processing of MM related tasks without the messages having to traverse the CN [113] –Provisions context awareness, as delay sensitive applications can access edge clouds whilst delay tolerant applications can still access services located in the core network [113,114] •Edge clouds Cons –Require dedicated infrastructure and appropriate placement [25,31,115] –Require fast service migration strategies to ensure seamless mobility [55] From these pros and cons as well as the preceding discussions, it is evident that the SDN based solutions satisfies parameter RL2 (allowing for seamless mobility), RL3 (through the provision of decentralized solutions), RL4 (through the ability to re-program paths in CN via orchestration of OF rules) and RL5 (through the ability to utilize network statistics for traffic steering with the CN) for the reliability criterion. For the flexibility criteria, the SDN based mechanisms satisfy the parameters FL1 (through the capability of orchestrating policies dependent on flow type, slice, etc.), FL3 (by allowing for CN based MM solutions that will work in synergy with the access network based solutions) and FL4 (through the global view of the network wherein a variety of parameters such as network load, QoS requirements, etc., are considered). In terms of scalability, SDN based solutions satisfy parameters SL1 to SL3 (given the ability to manage and steer traffic flows with the ability of having a distributed, hierarchical or centralized implementation) and SL4 (due to the possibility of having a decentralized configuration). The DMM based solutions, however only satisfy parameters RL2 (allowing for seamless handovers) and RL3 (due to the decentralized nature) in the reliability criterion. Further, for the flexibility criterion, DMM based solutions only satisfy parameter FL1, i.e., they only offer granularity of service by preventing any mobility anchor. It is noteworthy though that, from the scalability aspect DMM based solutions, like SDN based solutions, satisfy parameters SL1 to SL4, and for the same reasons. Lastly, for the edge cloud based solutions, parameters RL2 (allowing for seamless mobility through fast access to data/processing capabilities upon migration to the target network) and RL3 (allowing decentralization of MM based services) are satisfied for the reliability criterion. CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 71 For the flexibility criteria, parameters FL1 (due to the ability to provision services based on mobility and application profiles), FL3 (by allowing for MM methods at the edge network level in addition to the access and core network based solutions), FL4 (by provisioning processing capabilities for user association/BS selection services) and FL5 (by allowing for context awareness in data caching according to user mobility) are satisfied. Additionally, for the scalability criteria, parameters SL1 to SL4 are satisfied by the edge cloud solutions. The reason being, they allow for decentralization which can consequently permit better capability to manage connections and control messages due to increasing number of users. 3.3.2.2 Access Network Solutions As part of the analysis for the access network solutions, we firstly present the pros and cons for each mechanism discussed in Section 2.3.3.2, as follows: •Phantom Cell method Pros –Grants the ability to a UE to connect to multiple BSs simultaneously, thus also granting redundancy in physical layer connections [116] –As per reference [116] and our contribution [C1], it provisions the ability to allow per-flow and per-user granularity of service –Handover support at access network level [116] –Ease of implementation due to existing standards on MR-DC [45,116] •Phantom Cell method Cons –Handovers between different MC domains will still entail service disruption [45, 116] –According to [116] and our contribution [J1], Inter-MC domain handover signaling will still be a significant burden on the CN •RANaaS Pros –Provisions on-demand allocation of network resources at the RAN level [120–122] –Provisions the ability to execute on-demand handovers, through close interaction between the various RATs that are integrated at a BBU pool [123] –Assists in allowing UEs to camp on more than one BS –Introduces support for executing handovers at the access network level [123] CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 72 –Introduces the ability to utilize per-flow/channel granularity of service by being able to manage the physical connections more centrally [120–123] •RANaaS Cons –Requires a complete architectural overhaul at the RAN side of the network [120– 122] •Cross layer Pros –Allows for the sharing of network statistics between the various OSI layers [117– 119] –Allows for interaction between multiple OSI layers, thus facilitating the possibility of efficient utilization of multi-homing [67,117–119] •Cross layer Cons –Requires significant software modifications to the existing modular nature of the protocol structure [117–119] •Intelligent RAT selection Pros –Optimized RAT selection strategies [44,124–127] –Utilization of multiple parameters, such as BS load, UE context, etc., jointly for RAT selection [44,124–127] –Provisioning the ability to select RATs per-slice/user/flow [127] –As per our contribution [J3], it provisions the ability to select multiple BSs (possibly belonging to multiple RATs) •Intelligent RAT selection Cons –Requires rapid collection of network statistics to perform well informed selection –Based on our contribution [J3], computational complexity and convergence time of RAT selection algorithms will be critical, given the QoS requirements in 5G Given the discussions in Section 2.3.3.2 and the pros and cons listed above, we now determine the parameters, listed in Table 3.1, satisfied by each of the mechanisms explored. Concretely, for the phantom cell method, parameters RL1 (redundancy in physical layer connections) and RL2 (seamless mobility) are satisfied for the reliability criterion. For the CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 73 flexibility criterion, parameters FL1 (by permitting the possibility of per-flow and per-user based MM), FL2 (allowing for connectivity to multiple BSs potentially belonging to different RATs) and FL3 (provisioning handover support at the access network level that will work in synergy with CN based mechanism) are satisfied. In terms of scalability, the phantom cell method satisfies parameters SL1 to SL3 (owing to the handling of handover related computation and decision at the access network) and SL5 (owing to the existing standards on MR-DC, as discussed in Section 2.3.2). Next, the RAN-as-a-service concept satisfies parameters RL2 (allowing for seamless handovers) and RL5 (the softwarized nature enables dynamic initiation for RAN functionality such as BBU resources, functional splits, etc., depending on the network and user context) for reliability, parameters FL1 (allowing for per-flow, per-user, per-slice, etc., service granularity through its softwarized nature), FL2 (allowing the possibility for connecting a user to multiple BSs through its softwarized nature), FL3 (provisioning handover support at the access network which will work in synergy with the CN and edge network based methods) and FL4 (enabling the possibility of collection and utilization of RAN based information and generating intelligent BS selection/user association decisions) for flexibility, and parameters SL1 to SL3 (by offloading handover decision making and signaling to the access network) for scalability. On the other hand, cross-layer methods only satisfy parameters RL2 (allowing for seamless handover) and RL5 (allows for congestion aware method by sharing statistics about queue lengths, buffer sizes, etc., amongst the various layers) for the reliability criteria. Further, for the flexibility criteria they satisfy only parameters FL2 (by allowing for the possibility of multi-homing, etc.) and FL4 (allowing for the possibility of sharing statistics and other information amongst the various OSI layers and enabling joint optimization for BS selection, path re-routing, etc.). Lastly, for the intelligent RAT selection methods parameter RL2 (allowing for seamless handover through optimized decisions on RAT selection) is satisfied for the reliability criterion. For the flexibility criterion, parameters FL1 (allowing for the possibility of flow/user/slice based RAT selection), FL2 (allowing for the possibility to select multiple RATs for a given user) and FL5 (via the ability to utilize user and network context for RAT selection) are satisfied, while for scalability only parameter SL5 (owing to the extensive body of research for optimal RAT selection strategies) is satisfied. 3.3.2.3 Extreme Edge Network Solutions We firstly present the pros and cons for the D2D strategies as follows: CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 74 •D2D strategy Pros –Provisions D2D handover management strategies [128,129,132] –Provisions MM support at the extreme edge network level [128–132] –Provisions the ability to decentralize MM functionality •D2D Strategy Cons –Control signaling overhead will be a challenge [128,129] –The viability with regards to energy efficiency of D2D peers as well as latency incurred in conveying the decisions with regards to MM are un-explored questions Based on the discussions and the aforesaid pros and cons, the device-to-device methods satisfy parameter RL2 (through the provision of various seamless handover management studies) for reliability, parameter FL3 (provisioning mobility support at the edge network level which will work in synergy with access and core network based methods) for flexibility, and parameter SL4 (allowing for the decentralization of MM functionality) for scalability. 3.3.3 B5G Networks In this subsection we present a short study detailing the challenges that current state-ofthe-art mechanisms will continue to face for B5G networks. Furthermore, given the special characteristics that B5G networks will pose, as shown in Figure 2.1, we also list potential research areas for MM in B5G networks. Note that, these are then utilized in the subsequent section wherein we define challenges and potential solutions for 5G and beyond MM. Concretely, while SDN and NFV will provide the tools for the B5G networks to provision rapid programmability of the meta-surfaces, during mobility scenarios they will be challenged critically. The reason being that, while current networking paradigms permit anywhere between 1 ms–10 ms time interval for performing any programmability task (latency restrictions, as specified in current 5G networks [38], on most services), in B5G networks this will be constrained even further as additional surfaces need to be programmed and orchestrated. Specifically, an increased number of surfaces/network nodes leads to more data to be processed for generating appropriate programmability decisions. These decisions then need to be sent out (orchestrated) to the relatively large number of network nodes (including meta-surfaces), to execute the given task. Hence, this leads to an increased latency constraint on the network programmability aspect. Further, while the meta-surfaces provide a higher degree of freedom to the operator, they need to be programmed, as mentioned above. This CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 75 introduces the challenging aspect of managing the SDN domains, NFV orchestration and the related signaling. As a consequence, the compactness as well as the efficiency of the current state-of-the-art SDN and NFV procedures will be challenged. Next, with techniques such as DC, the challenge will be multi-fold as B5G networks will not just comprise of meta-surfaces, which can also act as a MIMO array, but they will also be equipped with Terahertz and mobile BS based multi-tier networks. And while, DC and multi-RAT procedures, as stated in Sections 3.3.1 and 3.3.2, will aid in ensuring a contextaware network selection procedure, the complexity for the access network techniques will be compounded by the fact that not only will they need to ensure QoS requirements, but they will have to also ensure sufficient available access bandwidth as well as backhaul bandwidth. Note that with the backhaul bandwidth there will be a significant design challenge since VLC technology is capable of carrying data rates of up to 1 Tbps. Current backhaul technologies cannot provision such high bandwidths [52]. Further, it is important to reiterate that the network will be composed of not only 4G-LTE and mmWave BSs, but there will also be VLC and drone based BSs, which essentially are the main reason for the increased complexity as discussed above. Moreover, for the edge clouds, while they aid in allowing low latency access to cached content as well as the compute resources, the deployment strategies will need to be rethought given the ongoing growth pattern for data usage as well as the number of served devices coupled with more resource hungry services. Certain important recent studies in this direction have been provisioned via references [140,141]. Given these significant shortcomings in the current state-of-the-art mechanisms towards B5G networks as well as taking into account the seminal works in the area of B5G techniques [32,33,35,36] [39], the potential areas of research in MM for these networks are as follows: •Characterization of the channel between meta-surface and the users, and meta-surface and the BS, in the event of user/BS being mobile, for the purpose of MM decisions •Consideration of reliability and coverage of VLC link for MM decisions •Characterization of the computational complexity for re-calibrating the meta-surfaces alongside the network, during mobility events •Impact of mobility upon the programmable environment1concept, drone based communication and VLC •Optimal RAT and BS selection with a programmable environment 1By environment, we refer to the physical environment that lies between the transmitter and receiver. CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 76 •Optimal RAT and BS selection in scenarios where both the UE and BS (drone based) are mobile •Characterizing the computational complexity of optimization methodologies for user association •Methods to handle possible increase in handover signaling/messaging during other network processes, such as reprogramming meta-surfaces to serve mobile users •Formulation of a sound heterogeneous RAT strategy, just like the 4G-5G concept, given mmWave and Terahertz technologies and their associated challenges related to coverage. Note that, the aforementioned research areas do not form an exhaustive list, but are broadly indicative of what aspects remain to be explored with regards to MM in B5G networks. To summarize, in this section we firstly introduced the 5G service based architecture and the classification of the various mechanisms that we analyzed, through Figure 2.13. Following this, we qualitatively analyzed the 3GPP 5G MM mechanisms as well as other research efforts with regards to their efficacy towards 5G and beyond MM solutions. Consequently, we introduce Table 3.3 wherein we indicate the parameters that each of the explored methods satisfies for the reliability, scalability and flexibility criteria (Table 3.1). We also enlist the important references that have lead us to the development of Table 3.3, as presented in this chapter. And so, from the capability profiles of each mechanism, as illustrated in Table 3.3, it is evident that even after significant efforts none of them completely meet the specified requirements as expected for the 5G and beyond MM mechanisms. Concretely, neither the 3GPP 5G MM mechanisms nor the other academic and industrial research efforts satisfy all the criteria completely. Subsequently, it is deduced that none of the analyzed mechanisms satisfy the requirements for the future MM mechanisms, as listed in Table 2.1. Hence, through the aforesaid qualitative analysis we have further exposed the gaps in the design and development for 5G and beyond MM mechanisms. CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 77 Table 3.3: Compliance with Reliability, Scalability and Flexibility criteria of Current state-of-the-art MM mechanism/standard 3GPP 5G MM SDN based DMM based Edge Clouds Phantom Cell RANaaS Cross layer Intel. RAT sel. D2D Cf.∗Refs.δCf. Refs. Cf. Refs. Cf. Refs. Cf. Refs. Cf. Refs. Cf. Refs. Cf. Refs. Cf. Refs. Reliability RL1 ! [45] × [106] × [24] × [25] ! [116] × [120] × [67] × [44] × [128] RL2 ! !!!!!!!! RL3 ![99] ![107] ![53] !× × × [117] ×[124] ×[129] RL4 ×[15, 102] ![108] × [22, 23, 110] ×[113– 115] × × [121, 122] ×[118, 119] ×[125– 127] ×[131, 132] RL5 ! ! [109] ×××! ! × × Flexibility FL1 ! [45] ! [106] ! [22] ! [25] ! [C1] ! [120] × [67] ! [124] × [128] FL2 !×××! ! ! ! × FL3 ! [15, 99, 100] !×!!! ×[117] ×![129, 130] FL4 ×[102] ![107] ×[53] ![113– 115] ×[116] ![121– 123] ![118, 119] ×[125– 127] ×[131, 132] FL5 !× × !× × × !× Scalability SL1 ! [45] ! [108] ! [53] ! [25] ! [45] ! [120] × [117] × [44] × [128] SL2 ×!!!!! ××× SL3 ×[15] !!!!! × × [124] ×[129, 130] SL4 ![100] ![109] ! [22, 23, 110] ![113– 115] ×[116] ×[121– 123] ×[118, 119] ×[125– 127] ![131, 132] SL5 !×××!× × !× ∗The conformance (Cf.) of a given mechanism for a given criterion. δThe corroborating references (Refs.), if any, for the specified conformance of a mechanism for a given criterion CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 78 3.4 Mobility Management: Persistent Challenges, Potential Solutions and Next-Generation Framework From our discussions in Chapters 2 and 3 so far, we have highlighted the requirements from MM mechanisms as well as the criteria that future MM mechanisms should satisfy to meet these requirements in Tables 2.1 and 3.1, respectively. Further, we have analyzed the legacy mechanisms and the current state of the art towards their utility for 5G and B5G networks in Tables 3.2 and 3.3, respectively. However, we have observed that gaps in fulfilling the requirements still persist. Concretely, we have demonstrated that none of the strategies evaluated satisfy the reliability, flexibility and scalability criteria in their entirety. Hence, to be able to design and develop a holistic MM mechanism, it is of substance to our study to understand the challenges/questions that persist. We consolidate, from earlier works in literature and the discussions in Chapter 2 and Sections 3.1-3.3, these key challenges/questions in the text that follows. 3.4.1 Challenges 3.4.1.1 Handover Signaling Even after the release of 3GPP specifications for 5G [29], HO signaling is still a challenge. Hence, reducing HO signaling to ensure system scalability and reliability will be one of the key challenges. Certain studies such as our contribution [J1], discussed in detail in Chapter 5, have provided methods to help overcome this challenge and thus, can be actively pursued by the research and industrial community. 3.4.1.2 Network Slicing Network slices have been defined to ensure different service types are served according to their own resource demands. Hence, it will be a key challenge to design MM strategies that either jointly take into account the requirements of multiple network slices or provide individual solutions for each network slice. 3.4.1.3 Integration framework for MM solutions The state of the art and 3GPP specifications ensure to some extent the provision of flexibility, reliability and scalability for 5G MM solutions, as discussed earlier. However, since these solutions function at different sections of the network (Figure 2.13), the challenge will CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 85 DMM (SDN-NFV integrated) Cross layer RAT selection D2D (with CP-DP extension) Edge Clouds Deep Learning Multi-homing On demand MM Smart CN Signaling SDN and Network Slicing Support NDN-ICN Core Network Level (SMF/Network Functions/AMF) Access Network Level (AMF/RAN/APMetasurface) Extreme Edge Network Level (devices) Handover Management RAN-as-aService MultiConnectivity Figure 3.1: Proposed 5G and beyond MM framework. Consequently, such methods together can provision more scalable, flexible and reliable MM strategies. And so, up until now in this section, we have highlighted the multiple challenges that the 5G and beyond MM mechanism will face, given our qualitative evaluation for legacy and current state-of-the art methods in Sections 3.1-3.3. We have then provisioned a brief discussion on the potential solutions that can assist in addressing these challenges. We illustrate a novel mapping between these challenges and potential solutions in Table 3.4. Additionally, we have also listed the parameters for the qualitative analysis (and hence the requirements specified in Table 2.1) that they satisfy. This, as a result, reinforces the completeness of our current study. Hence, in the next subsection, utilizing the inferences from Sections 3.1-3.3 and Table 3.4, we propose a framework for 5G and beyond MM. 3.4.3 Proposed 5G and beyond MM framework We utilize the earlier established classification process for the current state-of-the-art strategies to define our vision for 5G and beyond MM in Figure 3.1. Concretely, we have categorized the MM mechanisms as Core Network level,Access Network level and Extreme Edge Network level, depending on where they will be creating an impact on/from. The specific entities (based on the 5G architecture illustrated in Figure 2.13), to which these aforesaid levels correspond to, have also been mentioned in Figure 3.1. To elaborate, the core network strategies encompass the DMM, SDN and Network slicing paradigms to provision the necessary reliability, flexibility and scalability from a more global perspective. Additionally, the aforesaid core network strategies need to be well complemented with an efficient CN signaling strategy. Next, handover management, on-demand MM, IPv6 multi-homing and Edge cloud related MM strategies will be enacted not only in the core CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 86 network or the access network level, but jointly at both levels thus provisioning the necessary flexibility and reliability. Further, RAN-as-a-Service and Multi-connectivity provisions at the access network level will assist in utilizing the multiple RATs and BSs effectively. Moreover, it is envisioned that the RAT selection process maybe either at the access network or at the device level. The D2D techniques, on the other hand, are expected to provide added assistance for mobility at the device level through DP as well CP functionality. Complementing these mechanisms, NDN-ICN support will be provisioned at all levels, thus assisting in maintaining IP addresses/prefixes during mobility whilst resolving destinations via names. Note that, traditional IP address/prefix allocation strategies are not intended to be changed. Instead, the NDN-ICN concept provisions an over-the-top assistance. Further, the cross layer strategies, as the name suggests, will spawn across the multiple levels and enact policies, utilizing the available information at each of these levels, which assist in optimal MM related decisions across the network. Lastly, the deep learning strategies will again assist across the multiple levels by learning the complex features about the network context, user mobility and overall QoS requirements, and formulating effective MM related decisions. Hence, given that we utilize the potential solutions for overcoming the technology gap, specified in Section 3.4.2, alongside certain strategies from the state of the art and legacy MM mechanisms, specified in Sections 3.2 and 3.3, it can be inferred from Tables 3.1-3.4 that our proposed framework will satisfy all the parameters for the reliability, flexibility and scalability criteria. Consequently, it can be stated that the proposed framework in Figure 3.1 will also satisfy all the requirements as defined in Table 2.1, thus provisioning a holistic solution. With this vision, in the following section we summarize the developments of this chapter and establish how the work presented in subsequent chapters aims to realize this framework. 3.5 Summary In this chapter, we have provided a novel qualitative gap analysis of the legacy and current MM mechanisms with respect to their suitability for 5G and beyond networks. Given the complexity of future network scenarios, i.e., 5G and B5G, a full view of the MM strategies, their capabilities, the persistent challenges and the possible solutions to them, will enable the research community to design better MM strategies. Concretely, and in continuation with the requirements specified in Table 2.1 in Chapter 2, we firstly defined the three pillars of future MM strategies, i.e., scalability, flexibility and reliability, in-depth in Section 3.1. Further, we also specified the multiple parameters that the CHAPTER 3. QUALITATIVE GAP ANALYSIS IN MOBILITY MANAGEMENT 87 future MM mechanisms will need to satisfy for each of the evaluation criteria, through Table 3.1. Next, from our discussions in Section 3.2 it is clear that the legacy MM solutions fail in provisioning scalability, flexibility and reliability simultaneously. Nevertheless, the current standards and research efforts explored in Section 3.3 are promising as they provide enhanced capabilities towards future MM solutions. We have summarized these conclusions effectively in Tables 3.2 and 3.3. And as a consequence, through this qualitative analysis the various benefits and shortcomings of the legacy and the current state of the art mechanisms, studied in this chapter, can be understood easily by the research community. Subsequently, we established that none of the mechanisms fulfill the complete 5G and beyond MM mechanism requirements. And so, it is evident that a holistic MM mechanism for 5G and B5G networks remains elusive. Thus, certain challenges that will still persist for the design, development and deployment of future MM mechanisms have been detailed in this chapter in Section 3.4.1. Furthermore, we have provided a concise discussion on the potential MM strategies that the research community can explore so as to solve these persistent challenges and the technological gaps they present, in Section 3.4.2. Following this, we have also provisioned a novel mapping between the potential strategies and the persistent challenges in Table 3.4, thus highlighting the efficacy of our current study. Based on the inferences drawn, we have provisioned a novel framework for the 5G and beyond MM strategies through Section 3.4.3 and Figure 3.1. Henceforth, in Chapters 4-6, we build upon this future framework and provision an ondemand MM paradigm (Chapter 4); a novel Handover signaling mechanism (Chapter 5); and a novel User Association and Resource Allocation framework (Chapter 6), in order to facilitate enhanced MM solutions for 5G and beyond networks. Notably, each of these research works maps to one of the building blocks of the proposed framework presented in Section 3.4.3 and Figure 3.1. Specifically, the on-demand MM paradigm explored in Chapter 4, maps to the on-demand MM block in Figure 3.1, the novel handover signaling mechanism explored in Chapter 5, maps to the Smart CN signaling and Handover management blocks in Figure 3.1, and the novel User Association and Resource Allocation framework explored in Chapter 6, maps to the Multi-connectivity, RAT selection and Cross layer blocks in Figure 3.1. Chapter 4 Mobility Management as a Service Overview Mobility Management (MM) techniques have conventionally been centralized in nature, wherein a single network entity has been responsible for handling the mobility related tasks of the mobile nodes attached to the network. However, an exponential growth in network traffic and the number of users has ushered in the concept of providing on-demand mobility management, i.e., Mobility Management as a Service (MMaaS), to the wireless nodes attached to the 5G networks. Allowing for on-demand mobility management solutions will not only provide the network with the flexibility that it needs to accommodate the many different use cases that are to be served by future networks, but it will also provide the network with the scalability that is needed alongside the flexibility to serve future networks. And hence, in this chapter, a detailed study of MMaaS has been provided, highlighting its benefits and challenges for 5G networks as compared to the 3GPP, IEEE and IETF initiatives. Additionally, the very important property of granularity of service, which is deeply intertwined with the scalability and flexibility requirements of the future wireless networks, and a consequence of MMaaS, has also been discussed in detail. Contributions [C1] A. Jain, E. Lopez-Aguilera, and I. Demirkol, "Mobility Management as a Service for 5G Networks", IEEE ISWCS 2017, pp. 1–6. Existing mobility management architectures, such as the one employed by LTE [61], are centralized in nature. To illustrate, the Mobility Management Entity (MME) in the LTE architecture depicted in Figure 2.2 is the central entity which is entrusted with the respon88 CHAPTER 4. MOBILITY MANAGEMENT AS A SERVICE 89 sibility of managing mobility of users attached to the network. The aforementioned central architecture suffices current day needs. However, due to an exponential growth in traffic and the number of users, as stated in Chapters 1-3, these architectures will not be viable for the 5G network scenarios. Further, based on the discussions in Chapter 3, it is evident that factors such as lack of scalability and flexibility will render the current strategies insufficient for the scenarios that will prevail in these future networks. It is also important to state here that the scalability and flexibility of MM strategies is very intricately connected to the granularity of service aspect. In realization of the aforementioned issues, Software Defined Networking (SDN) and Network Function Virtualization (NFV) have been recognized as important enablers for the future networks. Concretely, they enable the implementation of critical network functions, such as mobility management, as applications on top of a central/distributed controller. And, with a global/locally-global perspective of the complete network architecture, the mobility management applications can be employed on an on-demand basis for the user devices. It must be mentioned here that the aforementioned global perspective relates to the scenario when the employed MM application has the complete network view, whilst a locally-global perspective indicates that the employed MM application has a global perspective of only a specific domain, which, for example, may be a geographical area that the SDN-controller, on which it is employed, covers. Further, the granularity of service provided by mobility management applications will also equip the networks with pre-requisites such as flexibility and scalability necessary to meet the demands of the 5G networks. Henceforth, in this chapter we discuss in detail the features of this on-demand mobility management service, better known as MMaaS, as well as the related granularity aspects along with their benefits and challenges for 5G networks. 4.1 MMaaS Existing mobility management strategies have primarily been centralized in nature. But, with the SDN and NFV techniques, mobility management functionalities can now be implemented as an application on top of a controller that provides it with a global view of the domain it serves. This softwarized control over mobility management permits the operators to provide the services on-demand, i.e., MMaaS. It is worth noting that, under the current mobility management strategies, when an MN attaches to the network, a mobility instance is created for it in the MME, and is kept at all times until it de-registers from the network. This leads to the unnecessary utilization of computational resources. However, with MMaaS, mobility management instances can be created on-demand and hence, computational resources CHAPTER 4. MOBILITY MANAGEMENT AS A SERVICE 90 can be allocated likewise. Consequently, MMaaS, through its global view and on-demand computational resource allocation, enables the provision of globally optimized solutions for managing user mobility. The aforementioned softwarized control allows for the utilization of a versatile set of parameters, which not only provide a globally optimal solution but also permit the selfadjustment of the established mobility management mechanisms. In order to retrieve these parameter values from the network entities, a network controller, i.e., the SDN-controller (SDN-C), has to interact with these entities over the southbound interface (SBI) and then pass on the extracted values to the mobility management application over the northbound interface (NBI) [153]. An illustration of the aforementioned process is provided through Figure 4.1. As can be seen from Figure 4.1(a), the SDN-C is connected to the OpenFlow (OF) switches, which comprise the network data plane. These switches are additionally also connected to the access network, from where values of the parameters such as Signal to Noise Ratio (SNR)/Received Signal Strength Indicator (RSSI) of other and current base stations (BSs) at the MN, types of flows on the MN, MN policies, etc., can be enquired. Further, from the OF switches, information related to the network such as network load, link failure/congestion information, as well as the latency over the links, etc., can be extracted. All of this information is then processed at the SDN-C which then, as is visible in Figure 4.1(a), is sent over the NBI to the mobility management application. These mobility management applications, which may be implemented on a software cloud, after processing this data, provide a solution to the SDN-C, which implements it over the network via the orchestrator through the SBI. Figure 4.1(b) provides a signaling diagram to illustrate the above flow of information to ensure mobility management services to the MNs attached to the network. It is important to mention here that the message sequence as provided in Figure 4.1(b) might change in practical implementation. However, the overall logical flow, i.e., information enquiry →information reception →information processing →MM rule implementation, is maintained. Further, in Figure 4.1(a), the access network might consist of a Centralized/Cloud RAN (C-RAN) [21], which is primarily composed of a BBU pool and multiple BSs attached to this BBU pool. In the aforementioned scenario, the BBU pool is responsible for handling the access network mobility, i.e., handling the mobility of MNs when they switch BSs within the same BBU domain. Here, a BBU domain specifically refers to a set of BSs controlled by a particular BBU pool. And so, as a consequence, the resource allocation rules message, as shown in Figure 4.1(b), is sent by the SDN-C to the access network only when the scenario demands, for instance: when performing a traffic transfer due to HO. Thus, it can also be inferred that MMaaS is essentially distributed wherein the access network mobility is handled at the BBU pool (or BS in case CRAN is not present), whilst network CHAPTER 4. MOBILITY MANAGEMENT AS A SERVICE 91 NBI SBI SBI SBI Mobility Management Applications, and other EPC functions SDN-C OF Switch OF Switch OF Switch Access Network BS BS BS BS (a) Enquire Parameters Parameter Values Parameter Values OF Rules Resource allocation rules Enquire Parameters Parameter Values Mobility Management Solution Access Network OF-Switch SDN-C Mobility Management Application (Software Core) Handover Required (b) Figure 4.1: (a) Softwarized network control; (b) Signaling diagram for mobility management in SDN based networks. level mobility (inter-domain mobility) is handled at the SDN-C. Lastly, an important point of consideration with regards to the resource allocation rules procedure is that, in the event legacy RAN deployments exist, the SDN-C, similar to the MME as shown in Figure 2.2, will almost always be in communication with the access network to handle the mobility at the access network level. Next, in order to further exemplify the advantages that MMaaS provides, a short comparative analysis with respect to the existent/legacy mechanisms has been provided in Table 4.1. The comparative analysis is conducted on the basis of 4 distinct parameters, i.e., Granularity of service,Degree of centralization,Network slicing support and Self-reorganizing capabilities. Note that, for this comparative analysis, we have utilized a broader umbrella of definition for the existing mobility strategies by considering them on the basis of the Standards Development Organization (SDO), i.e., IEEE, IETF and 3GPP, that were responsible for their development. Concretely, all the MM strategies discussed in Chapters 2 and 3, can be assimilated into these broad definitions without loss of specificity. And so, from Table 4.1, it is evident that the existing mobility management mechanisms, designed and developed by IEEE, IETF and 3GPP, do not provide a significant level of service granularity. Whilst these mechanisms provide at best per-MN granularity in service, MMaaS on the other hand has the ability to provide multiple avenues of granularity such as those based on mobility profiles, flows, policies, MN, etc. It is imperative to state here that, in order to provide the level of flexibility and scalability to the future networks, as CHAPTER 4. MOBILITY MANAGEMENT AS A SERVICE 92 would be needed to serve the complex scenarios that will be prevalent, such multi-avenue provision in granularity for mobility management services is an indispensable feature. In addition to the multiple avenues of granularity in service, MMaaS provides a significant advantage over the existing mechanisms by allowing for a de-centralized implementation of mobility management applications. Such flexibility stems from the softwarized control which is a consequence of the SDN and NFV framework. On the other hand, 3GPP due to the centralized MM anchors in the CP, i.e., the MME in 4G and SMF/AMF in 5G, and in DP, i.e., the S-GW in 4G and UPF in 5G, offers a very centralized strategy. However, through the LTE-X2/5G-Xn handover process and the traffic offloading mechanisms, some form of de-centralization can be obtained. IEEE mechanisms, such as 802.21 and 802.11x series, do not provision any specific methods for decentralization. Additionally, while IETF has certain studies on DMM, the overall architecture as defined by the PMIPv6, FMIPv6, etc., is fairly centralized. MMaaS through its softwarized control and global view will be able to serve multiple network slices whilst existing mechanisms, according to Table 4.1, cannot support network slicing environments. This is so because, existing mechanisms were not designed to logically slice the network infrastructure for the various tenants. By logically slicing the network infrastructure we mean that, resources in the core and access network are reserved independently for each tenant on it. Furthermore, by tenants we refer to services such as voice, broadband and Narrow-Band (NB) IoT, mobile virtual network operators (MVNOs), etc. It must be stated here that, 3GPP based mechanisms, due to the emergence of NB-IoT and already existing QoS request identifier mechanisms, provision certain level of network slicing support. Lastly, the self-organizing capabilities, i.e., the ability to re-structure the routing rules and the access network resource allocation (if needed) depending on the context of operation, are of prime importance to the future networks as the highly dynamic environment will require the mobility management mechanisms to adapt their solutions according to the scenario without any perceivable latency. MMaaS, owing to its flexibility and granularity characteristics as already mentioned, offers a high degree of self-organizing capabilities. On the contrary, existing solutions as shown in Table 4.1, offer minimal self-organizing features. Concretely, 3GPP, through LTE-X2/5G-Xn, LTE-DC/5G MR-DC and LWA, offers avenues for self-organization in the access network. IEEE, on the other hand, through the 802.21 and 802.11x suite, offers methods for resource allocation and negotiation in heterogeneous RATs. However, these functionalities are minimal as compared to MMaaS. Moreover, IETF, only provisions mechanisms that primarily function on the network layer and above. Hence, the self-organizing capabilities are only limited to provisioning support from the defined CHAPTER 4. MOBILITY MANAGEMENT AS A SERVICE 93 Table 4.1: Comparison between MMaaS and current/legacy architecture MMaaS 3GPP IEEE IETF Granularity of service Multiple avenues 1per-MN per-MN per-MN 2 Degree of Centralization Decentralized Mostly Centralized 3Centralized Centralized 4 Network slicing support Yes Minimal 5No No Self-reorganizing capabilities Very high Minimal 6Minimal Minimal mechanisms towards any self-organization rules/policies defined at the access network level. Thus, this discussion reinforces the belief that existing MM mechanisms are not wellsuited to handle the challenging scenarios that future networks will envisage. Furthermore, from the analysis so far, it is evident that granularity of service offers significant benefits to the network, through its provision of scalability and flexibility, as well as to the users, through the provision of optimal mobility management solutions dependent on their context. In addition, there are multiple avenues where granularity in mobility management services can be offered under the MMaaS concept. And so, in the subsequent section, a detailed study on granularity of service and the various avenues, such as mobility profiles, flow types, network load and policies, where the granularity can be offered has been provided. 4.2 Granularity of Service As stated in the previous section, granularity of service is beneficial to the network through its provision of scalability and flexibility, whilst for the users it formulates optimal mobility management solutions, which in turn benefit them by helping reduce the power consumption as well as improve their perceived Quality of Service (QoS). To better elaborate these afore1Per-flow, per-mobility profile, policy based, per-MN, etc. 2Multi-path TCP (MPTCP) and Stream Control Transmission Protocol (SCTP) allow for multiple paths/flows. However, IETF does not provide per-flow mobility management in these protocols as of yet. 3LTE-X2 and 5G-Xn handovers offer some form of de-centralization. 4IETF DMM working group presents certain studies on distributed frameworks. However, there are no standards level RFC as of now. 5Recently NB-IoT has been standardized, which utilizes LTE bands to serve IoT devices. 6LTE-X2/5G-Xn, LTE-DC/5G MR-DC and LWA allow some level of self-organizing capabilities. CHAPTER 4. MOBILITY MANAGEMENT AS A SERVICE 94 mentioned broad benefits, we consider the scenario as specified in Figure 4.2. The scenario specified is a typical mobility scenario wherein an MN migrates from one BS to another, and also switches its access router (AR) in the process. Further, at AR-1 the MN has a certain set of active flows. After moving from AR-1 to AR-2, the MN keeps its current flows active and also initiates other services which consequently result in the creation of new flows. It is imperative to note here that AR-1 and AR-2 are merely data-plane (DP) entities and henceforth, in the architecture mentioned in Figure 4.1(a), they are equivalent to the OF-switches. Next, whilst switching BSs and ARs, the mobility management application analyses parameters and policies which are relevant to both the user and the network. It also checks the context, i.e., the mobility profile, the flow types, etc., and makes a handover/traffic transfer decision. From the aforementioned decision process, it is clear that the mobility management rules implemented for the MN can provide granularity in terms of mobility profiles, flows, policy, and network load, or a combination of them depending on the context. As an example, in Figure 4.2, the granularity of service from the perspective of flows is illustrated, wherein delay-sensitive and delay tolerant flows are served individually with different MM rules. A more detailed discussion for the same is provided in Section 4.2.2. IP Core Network AR-1 AR-2 BS BS BS BS MN MN Delay sensitive flow (AR-1) Delay sensitive flow (AR-2) Delay tolerant flow (AR-1) Delay tolerant flow (AR-2) New flow (AR-2) Figure 4.2: MMaaS - Granularity of Service provision example. CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 101 ture with the Software Defined Networking (SDN) approach. As a result, we argue that the proposed mechanisms are viable and outperform the legacy handover signaling mechanisms in terms of latency incurred, total network occupation time, number of messages generated and total bytes transferred. Contributions [PT1] A. Jain, E. Lopez-Aguilera, and I. Demirkol, "Handover Method and System for 5G Networks", WO 2019/229219 A1 (WIPO PCT), pp. 1-98, 2019. (Positive International Search Report) [J1] A. Jain, E. Lopez-Aguilera, and I. Demirkol, "Evolutionary 4G/5G Network Architecture Assisted Efficient Handover Signaling", IEEE Access, vol. 7, pp. 256–283, Dec. 2018. (Quartile: Q1; IF: 4.098 (2018)) [C4] A. Jain, E. Lopez-Aguilera, and I. Demirkol, "Improved Handover Signaling for 5G Networks", IEEE 29th Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC) 2018, pp. 164–170. [C3] A. Jain, E. Lopez-Aguilera, and I. Demirkol, "Enhanced Handover Signaling through Integrated MME-SDN Controller Solution", IEEE 87th Vehicular Technology Conference VTC Spring 2018, pp. 1–7. Central to the solutions that offer seamless mobility support in wireless networks are the handover mechanisms, which allow a user to change its physical point of attachment within the network when it is subject to a mobility event and certain pre-programmed conditions are satisfied [133,156,157]. For example, if the RSSI or the received signal power from the current serving base station goes below a particular threshold and, simultaneously if the same parameter for another base station in the vicinity goes above a certain threshold, then a decision to change the point of attachment, i.e., the base station, can be taken by the network or the user. Further, given the highly heterogeneous scenario that will be prevalent in 5G networks, in this chapter we revisit the legacy handover mechanisms which form a critical part of MM. These legacy handover mechanisms are composed of four phases, i.e., handover decision (parameter values, such as RSSI, etc., based decision for BS selection), handover preparation (resource negotiation and allocation involving source and target networks), handover execution/rejection/cancel (path re-routing with the user transitioning from source to target network, or issuance of a cancel/reject indication due resource allocation failure) and CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 102 handover complete (release of source network resources upon successful handover to target network). Each of these stages contribute towards the overall latency and signaling cost to execute the handover. Hence, optimizing/enhancing them will facilitate in improving the QoE and QoS to the device/user. Consequently, many current research efforts, such as [44, 103, 146, 158–163], have provided studies and methods that will facilitate the enhancement/optimization of the aforesaid handover phases. However, the handover preparation and failure phases remain relatively unexplored in the studies referenced above and similar to them. It is during the handover preparation phase that the negotiation and allocation of resources for an impending HO is carried out. Further, during the handover failure phase, signaling that involves sending an indication to the source network and the user undergoing HO with regards to the failed HO attempt is performed. Thus, fast execution of the aforesaid signaling, in a markedly more complex network environment, will be a vital requirement for an efficient next-generation HO management framework. This requirement is further elaborated via the current and future network scenarios, and their corresponding HO phases, illustrated in Figure 5.1. In the current network scenarios, depicted in Figure 5.1(a), a user has significant time to trigger, prepare, execute and complete a handover. However, the same is not true for the future scenarios, as illustrated in Figure 5.1(b). In the future network scenarios, the density of base stations will be high, i.e., base stations with smaller coverage areas (Smallcells) and higher bandwidths will be packed more closely in a given area. In addition, Macro-cells with significantly large coverage areas will be existent as well, to assist the Small-cells. Hence, if current handover management strategies are utilized, the time taken to complete the handover will be much greater than the dwell time of the user (with its mobility profile) at the desired base station whilst the conditions are still favorable to establish a link. Specifically, the time available to perform the resource allocation and negotiation process will be shorter. Such a scenario would thus lead to loss of connectivity and hence, a poor network performance. Moreover, the HO signaling overhead will also be of critical importance for the network performance because of the FHOs caused by cell densification, and the diverse RATs used resulting in many inter-RAT HOs. Hence, an optimized handover process, where the HO latency and signaling overhead are reduced, will be an extremely vital component of future handover management strategies. Concretely, and from the discussion above, a fast and efficient handover preparation phase signaling will be important for an optimized handover process. Further, in the event that the HO has failed, the CN has to ensure a fast release of the allocated resources so that they can be reused, as well as the user under consideration is free to choose another base station. Thus the HO failure signaling process should also be optimized. CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 103 Macro Cell Macro Cell HO Execution HO Complete HO Required Connected to BS HO Latency HO Prep. UE (a) . . Macro Cell Macro Cell Small Cell Small Cell HO Prep. HO required Loss of Connection and QoS HO Latency Connected BS UE (b) Figure 5.1: a) Handover scenario in current wireless networks; b) Handover scenario in future wireless networks. Hence, in this chapter we have introduced a novel message mapping and parallelized control signaling methodology for the preparation and failure phase scenarios studied by the 3GPP [29, 133]. We follow this approach for both the legacy as well as the 5G networks. This approach subsequently results in a reduction of the overall transmission cost, processing cost, latency as well as the number of bytes transferred during a handover event. Further, the proposed approach has been designed not only for the Intra-RAT HO scenarios, but CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 104 also for the scenarios involving inter-RAT HOs including 5G HO scenarios discussed by 3GPP. In addition, in this chapter we have introduced a novel HO failure aware preparation phase signaling mechanism. The proposed mechanism accounts for the possibility of a HO failure during the design and execution of the HO preparation phase. And as will be seen in Section 5.3.3, the HO failure aware mechanism not only enhances the HO failure step but it also presents additional enhancement for the HO preparation signaling step. Further, in this chapter, we have provided a simple yet rigorous analytical methodology to validate the proposed mechanism. The aforementioned methodology enables the reader to compare the performance of the proposed mechanism with the legacy mechanisms, i.e., 3GPP standards, on the basis of latency incurred, transmission cost (i.e., total network occupation time), processing cost (number of messages generated) and the overall amount of bytes transferred. Note that, through the message size analysis we are able to determine the reduction in the overall amount of bytes transferred through the network during a given HO preparation or failure signaling sequence. It is important to state here that the packet or system level simulations would not be able to derive realistic network parameter values, since the network topology, the transport technology used, queueing at the network elements, etc. is dependent on the specific operator scenario and cannot be modeled accurately. Hence, we have utilized real data from network operators and vendors and have attempted to provide a simplistic, realistic, and yet holistic analysis. We have also introduced a novel network wide analysis. Through this we establish the fact that, given any distribution over the number of HOs for the studied HO types and for any HO failure rate, the proposed methodology greatly improves the system performance in terms of overall processing cost and total network occupation time, as compared to the legacy mechanisms. Next, the aforesaid message mapping and parallelized control information transfer is facilitated via an evolutionary network architecture. This network architecture establishes evolved CN entities wherein they are integrated with an SDN agent. Moreover, the MME in the 4G network and the SMF in the 5G network are evolved to SDN enabled CN entities. We refer to them as SeMMu. The reason for such an integration being that the MME and SMF are responsible for the CN signaling during a mobility event in their respective networks. Hence, this allows the SeMMu to facilitate the proposed HO signaling mechanism. Further, in this work, instead of the the AMF (which 3GPP defines as the mobility management unit in the 5G NGC), we exploit the idea of SDN-enabled SMF because it is the SMF which is involved in HO-related CN signaling, whereas the AMF is connected to the access network only. Thus, the HO-related CN signaling is not influenced by the AMF. Accordingly, in this chapter, we propose such evolutionary network architecture, by also describing the implementation aspects of such SeMMu entities. CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 105 Through this chapter, we propose an evolutionary architecture, considering the co-existence of the legacy networks (4G, 3G, 2G) and the newly proposed 5G networks by 3GPP, that enables a manageable CAPEX for the operators whilst also enhancing the handover preparation and failure phase performance. To summarize, in the current work we advance the state of the art by: •Introducing enhanced HO preparation and failure signaling phases for the myriad 5G and legacy networks inter-RAT HO scenarios. •Introducing a novel HO failure aware preparation phase signaling sequence. This approach will provision additional optimization to the legacy HO failure signaling step as well as the HO preparation phase. •Presenting performance analysis, based on latency, transmission cost and processing cost, of the proposed and legacy signaling mechanisms for the myriad 5G and legacy network HO scenarios specified by 3GPP. •Introducing a novel message size analysis for the proposed as well as legacy HO scenarios. •Introducing a novel network wide analysis in terms of the number of messages processed as well as the total network occupation time. •Presenting a novel 5G NGC and legacy inter-working architecture with and without the capabilities of an N26 interface. Note that, the N26 interface, defined by 3GPP, allows for the inter-working between the 5G and legacy networks. It allows for reduced signaling to prepare or reject a HO between 5G and LTE-EPC networks. •Presenting a novel interfacing mechanism between the MME/SMF and the SDN agent. 5.1 Legacy Handover Preparation and Failure Signaling Erstwhile standardization efforts by 3GPP [28,29,133,164–172] have led to the formulation of the handover signaling mechanisms currently being utilized in cellular networks [133] and to be used in future wireless networks [29]. Specifically, handover preparation and failure signaling phases will be critical to the overall system performance during mobility events. Figures 5.2, 5.3 and 5.4 illustrate the corresponding legacy handover preparation and failure signaling phases. CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 106 Figure 5.2: Legacy handover preparation signaling for Inter-RAT HO (5G-NGC to EPS) [29]. Figure 5.3: Legacy handover preparation signaling for Inter-RAT HO (LTE to 3G/2G) [133]. The legacy handover preparation signaling, exemplified in Figure 5.2 and 5.3, is initiated by a handover decision made by the source network. This is followed by a handover required message (#1 in Figure 5.2, and #2 in Figure 5.3). Following these initial stages, the handover preparation phase is comprised of resource negotiation and allocation through the RRM operations (messages 6 and 7 in Figure 5.2; messages 5 and 5a in Figure 5.3), as well as CP signaling to establish GTP tunnels. These GTP tunnels require the entities at either end of the tunnel to have the TEIDs and transport layer addresses of each other. Hence, the preparation step also encompasses the creation and exchange of TEIDs and transport layer addresses between the core network entities. In order to realize a successful handover preparation, handshakes between the core network entities, i.e., messages 4, 5, 8 and 10b in Figure 5.2; messages 4, 4a, 6, 6a, 8 and 8a in Figure 5.3, are required. Next, the legacy CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 107 Figure 5.4: Legacy handover failure signaling for Inter-RAT HO (5G NGC and EPS) [29]. handover failure phase signaling for inter-RAT HO (5G to EPS1) has been illustrated in Figure 5.4. For the 5G NGC, the HO failure phase signaling currently only includes the HO cancel mechanism. In 3GPP specifications, the handover failure phase signaling encompasses two different types of signaling methods, i.e., Handover Cancel and Handover Rejection. We define them as follows: •Handover Cancel: A handover cancel mechanism has been defined both for the 5G NGC as well as for the legacy networks, i.e., 4G, 3G and 2G. The cancel method is event based, i.e., it is initiated by a trigger event such as expiration of a timer, etc. It may be invoked only by the source network and at any point before the command to handover from the source network to the target network is sent from the MME/AMF to the source eNB/next generation NodeB (gNB). •Handover Reject: A handover reject mechanism is currently defined only for the legacy networks, i.e., 4G, 3G and 2G networks. Similar to the handover cancel phase, the handover rejection method is event based, i.e., it is triggered by an event such as failure to allocate sufficient resources at the target access network. Hence, upon reception of a rejected request to reserve resources, the target MME/SGSN informs the source eNB/RNC about the rejected requested and hence, a handover reject. And so based on the above definitions, due to certain network conditions, such as the expiration of a timer, etc., the source network may decide to cancel the HO (Figure 5.4). 1The EPS consists of EPC and E-UTRAN. Note that, the standard documents by 3GPP utilize EPS and EPC interchangeably while defining HO scenarios. Hence, in this chapter we utilize the same principle. CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 108 Consequently, the source MME/AMF informs the source base station with regards to the canceled handover (message 1 in Figure 5.4). Further, the source and target MME/SMF delete the sessions that had already been created with the target and source S-GWs/UPFs (messages 4, 4a, 7 and 8 in Figure 5.4), respectively. The creation and deletion of these sessions with other core network entities involves handshakes, which, as we will discuss in the following subsection (5.1.1), are a significant source of inefficiency in CN handover signaling. Note that the signaling schemes illustrated through Figures 5.2, 5.3 and 5.4 are representative and other HO preparation and failure phase scenarios explored by 3GPP in [29,133] are also of the same nature, wherein handshakes are utilized to accomplish the signaling procedures. 5.1.1 Signaling Inefficiency From our discussions and Figures 5.2, 5.3 and 5.4, it can be deduced that during the legacy handover preparation and failure phases, handshakes will be required to exchange the required CP information between the core network entities. For instance, in the handover preparation phase in Figure 5.2, to establish a session between the Target MME and the Target S-GW a handshake, i.e., messages 4 and 5, is required between these respective entities. Such handshakes, whilst being a reliable methodology, will occupy the network for a long period as opposed to a mechanism that does not involve any handshakes. Further, it will also lead to higher latency, signaling cost, processing cost and total bytes of data transferred. And given the future network scenario depicted in Figur 5.1(b), wherein the network will be dense and heterogeneous, the legacy mechanisms will be rendered inefficient. Thus, we define a new principle that is utilized to create a compressed message ensemble and an enhanced signaling method, showing increased performance in terms of latency, signaling cost, processing cost and total amount of bytes transferred. The principle is as follows: “Identify the sequence of messages, such as the handshakes, where the performance of the 3GPP defined methods can be improved/enhanced. Then re-shuffle the information elements (IEs), if possible, to form a compressed message ensemble such that the sequence of messages under scrutiny are executed efficiently, if possible in parallel, but with the desired functionality." And so, we next discuss the novel message mapping and signaling strategies that alleviate the deficiencies mentioned above. CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 109 Table 5.1: Different handover scenarios analyzed Target Network 5G NGC†EPC 3G/2G Source Network 5G NGC†N2 based HO: UE migrates from one NGRAN to another Inter-RAT HO involving an N26 interface Inter-RAT HO without an N26 interface EPC Inter-RAT HO involving an N26 interface Intra-RAT HO involving MME relocation but no S-GW relocation Inter-RAT HO involving S-GW relocation and an Indirect tunnel Inter-RAT HO involving S-GW relocation and a Direct tunnel Inter-RAT HO without an N26 interface Intra-RAT HO involving MME and SGW relocation Inter-RAT HO without any S-GW relocation but with an Indirect tunnel Inter-RAT HO without any S-GW relocation but with a Direct tunnel 3G/2G Inter-RAT HO involving an S-GW relocation and Indirect tunneling Inter-RAT HO without any S-GW relocation but with Indirect tunneling †3GPP standards document only discuss 5G NGC to EPS HO and vice versa. 5.2 Proposed Handover Preparation and Failure Signaling The proposed handover preparation and failure phases consist of a novel message mapping and signaling mechanism, wherein a compact and intelligent mapping of IEs from the legacy to the proposed signaling messages has been provided. Additionally, the proposed mechanism CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 110 also involves parallel transfer of CP information. To facilitate these capabilities, we utilize the SDN enabled CN entities including the SeMMu, defined earlier. Specifically, the SeMMu through its SDN capabilities and centralized location facilitates: •Parallel transfer of CP information to other CN entities. •Allocation of TEIDs and transport layer addresses for the other CN entities at the SeMMu itself. Thus, through the use of compact message ensemble and parallelization of information transfer, the handover preparation and rejection signaling for the various 3GPP HO types are, as we will discuss in this section, optimized. The HO scenarios that have been analyzed in this work are presented in Table 5.1. Concretely, the various networks, i.e., 2G, 3G, 4G-LTE and 5G, have been considered in this table and all the possible HO scenarios among them have been enlisted. To evince the optimization achieved for the aforementioned 3GPP HO scenarios, we consider a representative HO scenario, i.e., Inter-RAT HO from 5G NGC to EPS network wherein the serving gateway is relocated. The optimized/enhanced message maps and signaling sequence for other scenarios have been illustrated in the Appendix A. And so, for the representative scenario considered, a user undergoes an Inter-RAT handover with the source system being 5G and the target system being an EPS network. Further, serving gateway relocation defines that during the handover process the gateway that is serving the user is changed. In this particular scenario, the gateway in the source network is a UPF which upon handover to the EPS is switched to a target S-GW. Also, note that the considered scenario consists of an N26 interface which facilitates the inter-working between the NGC and EPS. However, in the analysis, we have presented results for the scenarios wherein the N26 interface does not exist. Such a scenario will be prevalent in the initial 5G standalone (SA) deployments, i.e., an interface between 5G NGC and EPS will be missing in some of the initial deployments, due to cost/compatibility reasons. Subsequently, we also provision the proposed HO signaling diagrams for this scenario in Appendix A. Note that, in addition to the proposed signaling method, in this section we also present a novel handover failure aware handover preparation method. This novel approach enhances not only the handover failure method but also improves the handover preparation method further. We defer the detailed discussion on this method to Section 5.2.3. CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 117 With this background, and through Figures 5.6, 5.7 and 5.10, it must be noted that if the HO cancel phase is executed before message P5a (Figure 5.6) is sent, then it will be fully optimal, i.e., there will be no requirement for handshakes to delete the tunnels. This is so because, there will be no CN resources that would have been allocated during the HO preparation steps at that point, and hence, there will be no requirement of messages P4a, P4c and P7 (Figure 5.7) to release those resources. However, if the HO cancel phase is executed at any time instance after the execution of message P5a (Figure 5.6) within the HO preparation phase, then the cancel phase signaling will require messages that will help release the allocated CN resources. Thus, the proposed HO cancel phase signaling presented in Figure 5.7 is capable of adapting itself depending on when it is initiated, and as a consequence we consider it to be near-optimal. UE E-UTRAN NG-RAN AMF T-SeMMu S-GW PGW-C + SeMMu PGW-U + UPF P1. Handover required Handover Initiation P7a. Indirect Data Forwarding Tunnel P8. Nsmf_PDUSession_UpdateSMContext Request P5. Handover Request P6. Handover Request Acknowledge P10(12). HO Command P9(a). N4 Session Modification Handover Decision P2. Nsmf_PDUSession_Context Request P4. Relocation Request P7b. Relocation Response P10. N4 Session Modification P9b(11). Nsmf_PDUSession_UpdateSMContext Response P3. Nsmf_PDUSession_Context Response P11(13). HO Command Figure 5.9: Handover failure aware Handover preparation Signaling for Inter-RAT HO from 5G NGC to EPS. Upon performing a deeper analysis into the IEs that constitute messages P5a, P7a and P9a (Figure 5.6), we observe that the IEs of messages P5a and P7a can be re-shuffled and mapped to create a single message P7a, i.e., Indirect Data forwarding tunnel, as shown in the Handover failure aware preparation phase signaling (Figure 5.9). Further, the new message P7a is executed after messages P5 and P6 (Figure 5.9), which facilitates the HO cancel phase with a greater chance of achieving its optimal state presented in Figure 5.10. The reason being, the later the tunnels are setup, the higher is the chance of a HO cancel event not CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 118 UE S-RAN T-RAN S-AMF T-SeMMu S-(PGW-C/ SeMMu) T-SGW PGW-U + UPF P1. Handover cancel P5. HO Cancel Acknowledgement Source RAN decides to cancel HO P3. S1 Release Procedure P2. Relocation Cancel Request P4. Relocation Cancel Response Figure 5.10: Optimal proposed Handover rejection phase signaling sequence for Inter-RAT HO from 5G NGC to EPS. UE Source eNB Target RNC SeMMu Target SGSN Source SGW Target SGW PDN-GW HSS P6a. S-SGW Tunnel setup P6b. T-SGW relocation and Tunnel Setup P4. Relocation Request P4a. Relocation Request Acknowledge P1. Handover Initiation P3. Resource Allocation Request + Tunnel setup P2. Handover required P5. Forward Reloc. Resp. P6c. Handover Command P7. HO from E-UTRAN command Figure 5.11: Handover failure aware Handover preparation Signaling for Inter-RAT HO from LTE-EPC to 3G/2G when there is indirect tunneling and S-GW relocation occurs. UE Source eNB Target RNC SeMMu Target SGSN Source SGW Target SGW PDN-GW HSS P4. Relocation Request P4a. Relocation Failure P1. Handover Initiation P3. Resource Allocation Request + Tunnel setup P2. Handover required P5. Forward Reloc. Resp. (Reject) P6. Handover Preparation Failure Figure 5.12: Optimal proposed Handover rejection phase signaling sequence for Inter-RAT HO from LTE-EPC to 3G/2G when there is indirect tunneling and S-GW relocation occurs. requiring to delete these tunnels as it may be initiated before they are setup. Additionally, CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 119 and in the event, the HO cancel phase is executed after message P9a (Figures 5.6 and 5.9), the aforesaid enhancement would require the HO cancel signaling to delete 2 tunnels (created through messages P7a and P9a in Figure 5.9) instead of the 3 (created through messages P5a, P7a and P9a in Figure 5.6). However, given the dynamic nature of HO cancel phase, in the analysis we only consider the enhanced signaling specified in Figure 5.7. This is also the worst case HO cancel phase scenario as it is executed after all the tunnels have been setup. Additionally, utilizing this novel signaling approach, the handover rejection phase for the LTE to 3G handover scenario, illustrated in Figure 5.8, has been further enhanced (Figure 5.12). To achieve this enhancement, the HO preparation phase for the given scenario is first modified (Figure 5.11) such that all the tunnel and session creation messages are executed after Relocation Request Acknowledge message (5a in Figure 5.3). The reason being, in the event there is a HO rejection, the Relocation Request Acknowledge message issues an indication of the rejection to the SGSN. The SGSN then passes this indication to the SeMMu, which instantly passes the reject indication to the source eNB, without the requirement of any session and tunnel deletion messages. Hence, the HO rejection phase signaling is further enhanced as compared to signaling proposed in Figure 5.8. Further, given that the resource allocation process is successful and a positive indication is received from the Relocation Request Acknowledge message, the tunnel and session creation messages are executed simultaneously with the HO command message. This parallel execution with the HO command grants additional enhancement to the HO preparation phase as it helps to reduce the latency further. 5.2.4 Xn, X2 and S1 Interface based Handover Signaling As an evolution from the EPC architecture, the 5G NGC specifies an Xn interface between two gNBs. In scenarios, wherein the two gNBs involved in the HO process are connected via an Xn interface, the given interface facilitates a faster handover process. Subsequently, upon deeper exploration of the handover preparation signaling mechanism for an Xn based HO from [173], it is evident that the existing mechanism is optimal. Concretely, since the signaling does not involve any handshakes and any significant interaction with the CN entities, the proposed handover preparation signaling process will neither provide any gains nor will it lead to any regressive effects on the performance of the Xn based HO mechanism. Further, the LTE-EPC defines two specific interfaces through which the Intra-RAT handovers can be executed, i.e., the X2 and the S1 interface [133]. Whilst, the X2 interface is defined between two eNBs (like the Xn between two gNBs in 5G NGC), the S1 interface involves the CN. The legacy X2 and S1 handover preparation signaling mechanisms have CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 120 been presented in [174]. Analyzing the signaling mechanisms presented in [174], it can be concluded that the existing X2 and S1 handover mechanisms, similar to the Xn handover mechanism for 5G NGC, are optimal. Concretely, while the X2 HO is similar to the Xn HO in 5G NGC (wherein the HO signaling is only at the access network), for the S1 HO the only CN signaling is that between the MME (SeMMu) and the source and target eNBs (wherein RRM operations take place). Hence, both X2 and S1 HO signaling scenarios are considered to be optimal, as stated above. Note that, here for the S1 handover we consider the scenario wherein the user does not switch its MME (SeMMu) and S-GW. In the event either is changed, the proposed handover mechanism leads to immediate gains, which have been presented via the analysis in Section 5.3. Additionally, it is important to state that the Xn, X2 and S1 interfaces are an integral part of the evolutionary network architecture proposed in this work, and are agnostic to the implementation of the architecture and signaling methodology. Thus, with the aforesaid principles, processes and methodologies, in the following section we present a detailed quantitative performance improvement analysis for the myriad scenarios that have been listed in Table 5.1. 5.3 Performance Analysis To analyze the proposed handover preparation and failure signaling phases, we use latency, processing cost, transmission cost and amount of bytes transferred within CN as evaluation metrics. Note that, as in [175], utilizing these metrics for evaluating new handover strategies is standard practice. 5.3.1 Analytical Formulation For the analysis we first define a set S={s1, ..., sN}corresponding to all the link delays encountered within a given signaling sequence. Then the set J={j1, ..., jK}is the set of all the parallel link delays, where K≤N. By Parallel link delay we mean that, if xmessages are to be executed simultaneously then the overall delay incurred will be the maximum of the delays experienced by the messages under observation. It is then computed as Parallel Link Delay =max(Link delay msg 1, . . . , Link delay msg x),(5.1) CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 121 Additionally, we consider only a single processing delay for the group of messages that are being executed in parallel. Hence, the set of processing delays can be defined as D={d1, ..., dK}. It is imperative to state here that the assumption for the aforesaid processing delays, mentioned in Section 5.3.2, is conservative in nature. Hence, any SDN agent processing delay is also included within the utilized assumptions. Also, in HO procedures no routing table updates would be necessary, as they are already configured during the network setup phase, hence no signaling overhead is created by SDN agents for HOs. And so, for the computation of latency incurred during the handover preparation and handover failure signaling phases, we consider the contributions from parallel link delays and processing delays as: Latency = K Õ i=1 {ji+di}(5.2) The Transmission Cost computation, on the other hand, requires that each link delay be considered for the evaluation, and is computed as Transmission Cost = N Í l=1 sl 1ms .(5.3) Concretely, the Transmission Cost analysis represents the amount of time the CN links are occupied during the complete HO signaling process. Next, for the processing cost analysis, we utilize the analytical methodology in [134] and define it as the number of messages generated during the HO preparation/failure signaling phase. We then compute the percentage processing cost saving as Proc. Cost Saving = MSGLegacy−MSGProposed MSGLegacy ∗100%,(5.4) wherein, MSGLegacyis the number of messages in the legacy approach for HO preparation/failure and MSGProposed is the number of messages in the proposed approach for HO preparation/failure. In addition, in this chapter we also present an analysis for the network wide processing cost and occupation time. Whilst the network wide processing cost will reflect the network wide reduction in the processing cost through the proposed method, the network occupation time will be reflective of the reduction in the amount of time the CN links are occupied due to the handover signaling across the network. To conduct the aforesaid analysis we introduce the formulations in equations (5.5) and (5.6). NPCDHfS1pand NOcTDHfS1pare the Network wide processing cost and Network occu- CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 122 pation time, respectively, given a HO distribution (distribution of percentage of total users undergoing a particular handover), HO failure (rejection/cancellation) rate and percentage of users undergoing S1 HO, respectively. Note that, we consider only the Intra-MME/S-GW scenario for S1 HOs as it is not impacted by the implementation of the proposed mechanism but it still involves CN signaling. Moreover, we do not consider the X2 and Xn handovers for the analysis, as they do not involve any significant CN signaling. The rest of the notations in (5.5) and (5.6) are as follows: Hpscost is Handover preparation processing cost vector; DistHO is the handover distribution vector; HOsperc is the handover success percentage; Hfcost is the processing cost vector during Handover failure; HO f perc is the HO failure percentage; NHO is the number of users undergoing handover in the network; S1pis the percentage of S1 handovers (Intra-MME/S-GW); S1scost is the processing cost for a successful S1 HO preparation; S1f cost is the processing cost for a failed S1 HO; Htscost is the transmission cost vector for a successful HO preparation; Htsfcost is the transmission cost vector during a HO failure scenario; S1tscost is the transmission cost for a successful S1 HO preparation; and S1ts f cost is the transmission cost incurred when a S1 HO fails. 5.3.2 Parameter Specification and Assumptions As part of the analytical framework, the parameter values that will be utilized to conduct the analysis are provided in this subsection. Firstly, the one-way delays for each CN link, necessary for the latency and transmission cost analysis, have been defined in Tables 5.2 and 5.3 by utilizing the data from a Japanese cellular operator [176], Cisco [177] and a Greek cellular operator. Further, the delays presented for each link are considered to be symmetric, i.e., if a delay of 1ms is incurred for the link from AMF to SeMMu, then the same link delay is assumed from SeMMu to AMF. In Table 5.2 the link delays presented are derived from a Japanese operator deployment data [176] and CISCO data [177]. In addition, the link delays are computed considering NPCDHfS1p=n(Hpscost ∗DistHOT) ∗ HOsperc +(Hfcost ∗DistHOT) ∗ HOf per c o∗ (1−S1p) ∗ NHO+, S1p∗NHO ∗nS1scost ∗HOsperc +S1f cost ∗HOf per c o(5.5) NOcTDHfS1p=n(Htscost ∗DistHOT) ∗ HOsperc +(Htsfcost ∗DistHOT) ∗ HOf per c o∗ (1−S1p) ∗ NHO+, S1p∗NHO ∗nS1tscost ∗HOsperc +S1ts f cost ∗HOf per c o(5.6) CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 123 Table 5.2: Link Type and Corresponding Delays in Proposed Architecture (Derived from a Japanese Operator [176] and Cisco data [177] Link Type Link Delay 1. UE to NG-RAN 1ms 2. NG-RAN to AMF 7.5ms 3. AMF to SeMMu (PGW-C + SMF) 1ms 4. AMF to SeMMu 1ms 5. SeMMu to S-GW 7.5ms 6. SeMMu (PGW-C + SMF) to PGW-U + UPF 7.5ms 7. SeMMu (PGW-C + SMF) to PCRF+PCF 7.5ms 8. AMF to AMF 15ms 9. SeMMu to PGW 7.5ms 10. SeMMu to E-UTRAN 7.5ms 11. E-UTRAN to UE 1ms 12. PGW to PCRF 7.5ms 13. S-GW to PGW 7.5ms 14. SeMMu to SGSN 1ms 15. SGSN to RNC 6ms 16. SGSN to S-GW 7.5ms 17. SeMMu to SeMMu 15ms that the MME (SeMMu in this study) and the SGSN are co-located, as specified in [178]. Utilizing this co-location principle, we also establish the link latency between AMF and SeMMu. Further, the 15 ms SeMMu-SeMMu and AMF-AMF delay is based on the premise that the delay between the SeMMus/AMFs will be greater than the largest CN delay within a SeMMu/AMF domain. Hence, for the purpose of analysis in this chapter and for the data provided from the Japanese operator and Cisco, an assumption of two times the greatest link delay within a SeMMU/an AMF domain has been considered. On the other hand, the values of delays obtained from the Greek operator (Table 5.3) correspond to eNBs from two different networks and CN elements from 3 different MME domains. Consequently, for the chosen network and its MME domain, the link delays are computed as the average of all the delay values provided by the network operator for that specific link. Further, the UE-eNB and the eNB-SeMMu delay for both data sets is derived from the Cisco framework in [177]. Additionally, for the latency analysis, we consider the processing delay to be 4 ms in all CN entities, as in [177]. For the network wide analysis, we consider that the number of users undergoing handover CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 124 Table 5.3: Link Type and Corresponding Delays in Proposed Architecture (Derived from a Greek Operator and Cisco data [177]) Link Type Link Delay 1. UE to NG-RAN 1ms 2. NG-RAN to AMF 19ms 3. AMF to SeMMu (PGW-C + SMF) 0.5ms 4. AMF to SeMMu 0.5ms 5. SeMMu to S-GW 1ms 6. SeMMu (PGW-C + SMF) to PGW-U + UPF 1ms 7. SeMMu (PGW-C + SMF) to PCRF+PCF 1ms 8. AMF to AMF 2ms 9. SeMMu to PGW 1ms 10. SeMMu to E-UTRAN 19ms 11. E-UTRAN to UE 1ms 12. PGW to PCRF 1ms 13. S-GW to PGW 1ms 14. SeMMu to SGSN 0.5ms 15. SGSN to RNC 2ms 16. SGSN to S-GW 1ms 17. SeMMu to SeMMu 2ms at any given time in the considered network, i.e., the parameter NHO in (5.5) and (5.6), is 3 million. The analysis does not take into consideration the users that undergo an X2 or Xn based handover, i.e., they are not included amongst the 3 million users that we include in our analysis, as they do not involve any HO-related CN signaling. In addition, and based on discussions in Sections 5.2.2 and 5.2.3, the HO cancel phase is considered only for the 5G networks, while for the legacy networks (4G/3G/2G) we only consider the HO rejection phase signaling. Recall that, for the 5G networks the rejection phase signaling does not exist. Further, the considered HO cancel phase for the 5G NGC is as shown in Figure 5.7, which is also the worst case enhanced signaling for the same. However, for the legacy networks (4G/3G/2G) we do not consider the HO cancel phase since: •The HO cancel signaling process for the legacy networks is fundamentally the same as that in the 5G NGC. Hence, considering the HO rejection signaling phase for the legacy networks aids in the completeness of analysis and study. •Given the dynamic nature of HO cancel phase (Section 5.2), considering the HO rejection phase signaling also facilitates the ease of analysis. CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 125 We then develop five randomly distributed settings over the HO types (Table 5.1) for the computation of network wide processing cost and network occupation time. Concretely, we define the HO distributions that will be utilized for the analysis through equations (5.5) and (5.6), i.e., the parameter DistHOT. The distributions are generated using Algorithm 1, wherein one of the distributions is predefined to be uniform across the HO types. Through uniform we mean that the percentage of users experiencing a particular handover scenario is the same for all HO types. It is imperative to state here that, the premise behind considering random distributions over the HO types is the lack of availability of real data from network operators. Algorithm 1 Distribution Generation 1: procedure DistributionGenerator 2: iter ←5 3: i←1 4: mprct ←0.2 5: NoH ←Number of Handover Types 6: for i<iter do 7: maxper ←mprct 8: minper ←10−4 9: j←1 10: for j<=NoH do 11: Distper(i,j) ← U[minper,maxper] 12: maxper ←min(1−sum(Distper(i,:)),mprct) 13: j←j+1 14: Distper(5,:) ← ones(1,NoH)/NoH And so, in Algorithm 1 we first define the maximum percentage of users (maxper) that undergo a particular HO type to be 20%, whereas the minimum percentage (minper) of users that undergo a particular HO type is 0.01%. Next, to generate the random distribution, we utilize the uniform probability distribution (U), with its upper and lower bounds being specified by the maximum and minimum percentage, respectively. We continually update the maximum percentage so as to prevent any skewness in the nature of distribution. The update rule is defined as the minimum value amongst 20% (initial maximum percentage value) and the percentage of users that remain to be associated to a particular HO type. We then define the last distribution as being uniform across all the HO types (Algorithm 1: Line 14). CHAPTER 5. ENHANCED HANDOVER SIGNALING METHOD AND SYSTEM 126 5.3.3 Performance Analysis In this section, utilizing the formulation presented in Section 5.3.1, we present and discuss the analytical results for the latency, processing cost and transmission cost of the new signaling framework for the handover preparation and failure phases presented in Section 5.2. For the analysis, we utilize the link latency data shown in Tables 5.2 and 5.3. The analytical methodology undertaken here is used to compare the performances of the proposed approach and the current 3GPP defined approach. 5.3.3.1 Latency analysis Utilizing equation (5.2), as well as the cellular operator data from Section 5.3.2, we present the analytical results for the latency improvement for the handover preparation phase in Tables 5.4 and 5.5. Table 5.4: Preparation Phase: Handover Latency Improvement Analysis (Cisco and Cellular Operator-Japan) Handover Type Legacy Mechanism Proposed Mechanism Percentage Latency Reduction 1.Uρ155 ms 95 ms 38.71% 1.U4138.5 ms 31.41% 1.Vρ181 ms 89 ms 50.82% 1.V4171.5 ms 138.5 ms 19.24% 1.W 179 ms 123 ms 31.28% 1.X.a† 128 ms 65.5 ms 48.83% 1.X.b† 1.Y.a† 82 ms 65.5 ms 20.12% 1.Y.b†58 ms 29.27% 1.X.a∗129.5 ms 65.5 ms 49.42% 1.Y.a∗82 ms 65.5 ms 20.12% 2.y 113 ms 90 ms 20.35% 2.x 159 ms 90 ms 43.40% 1: Inter-RAT HO; 2: Intra-RAT (LTE) HO; a: Indirect Tunnel; b: Direct Tunnel; U: with N26 interface V: without N26 interface; X: with T-SGW; Y: without T-SGW; ρ5GS to EPS; 4EPS to 5GS y: inter-MME and intra-SGW; x: inter-MME and S-GW; ∗3G/2G to LTE; †LTE to 3G/2G W: Intra-NG-RAN N2 based HO in 5G NGC We show through this analysis that the proposed mechanism reduces the latency as compared to the legacy mechanism for both sets of operator data and all HO types considered. Note that, while the proposed mechanism helps reduce the latency by more than 19% for all HO types over the Japanese operator data (Table 5.4), the latency reduction over the Greek operator data (Table 5.5) ranges from 8.3% to 35.80%. Such differential behavior is 1. Handover Initiation 2. Handover Required 3. Forward Relocation Request 4. Create Session Request 4a. Create Session Response 5. Relocation Request 5a. Relocation Request Acknowledge 6. Indirect Data Forwarding Tunnel Request: T-SGW 6a. Indirect Data Forwarding Tunnel Response: T-SGW 7. Forward Relocation Response 8. Indirect Data forwarding Tunnel Request: S-SGW 8a. Indirect Data forwarding Tunnel Response: S-SGW 9. Handover Command 10. HO from UTRAN Command P1. Handover Initiation P2. Handover Required P3. Forward Relocation Request P5b. S-SGW Tunnel Setup P5a. T-SGW Relocation and Tunnel Setup P4. Relocation Request P4a. Relocation Request Response P5c. Forward Reloc. Response P6. Handover Command P7. HO from UTRAN Command Handover Preparation Handover Execution Complete Map Partial Map Message Flow Current HO Preparation Signaling Proposed HO Preparation Signaling Figure A.12: Proposed Handover Signal mapping for 3G/2G to LTE Inter-RAT HO with Target S-GW.. 229 UE Source e-NB Target e-NB Source SeMMu Target SeMMu Source S-GW Target S-GW PDN-GW HSS P4. Handover Request P5. HO request Acknowledgement P1. Handover Initiation P3. Fwd. Relocation Request P2. Handover required P6b. Forward Reloc. Resp. P7b. Handover Command P8. Handover command to UE P6a. T-SGW Indirect Tunnel Setup for forwarding P7a. S-SGW Indirect Tunnel Setup for forwarding Figure A.13: Proposed Handover Signaling for LTE Intra-RAT HO with Target S-GW and MME. 230 1. Handover Initiation 2. Handover Required 3. Forward Relocation Request 4. Handover Request 4a. Handover Request Acknowledge 5. Forward Relocation Response 6. Indirect Data forwarding Tunnel Request: S-SGW 6a. Indirect Data forwarding Tunnel Response: S-SGW 7. Handover Command 7a. HO Command to UE P1. Handover Initiation P2. Handover Required P3. Forward Relocation Request P4. Handover Request P4a. Handover Request Acknowledge P5. Forward Reloc. Response P6b. Handover Command P7. HO Command to UE Complete Map Partial Map Message Flow Current HO Preparation Signaling Proposed HO Preparation Signaling P6a. S-SGW Tunnel Setup Handover Preparation Handover Execution Figure A.14: Proposed Handover Signal mapping for LTE Intra-RAT HO with MME relocation (without S-GW relocation). 231 UE Source e-NB Target RNC SeMMu Target SGSN Source S-GW Target S-GW PDN-GW HSS P6a. S-SGW Tunnel setup P4. Relocation Request P4a. Relocation Request Response P1. Handover Initiation P3. Forward Relocation Request P2. Handover required P5. Forward Reloc. Resp. P6b. Handover Command P7. HO from E-UTRAN command Figure A.15: Proposed Handover Signaling for LTE to 3G/2G Inter-RAT HO without Target SGW and Direct Tunnel. 232 1. Handover Initiation 2. Handover Required 3. Forward Relocation Request 4. Relocation Request 4a. Relocation Request Acknowledge 5. Forward Relocation Response 6. Indirect Data forwarding Tunnel Request: S-SGW 6a. Indirect Data forwarding Tunnel Response: S-SGW 7. Handover Command 8. HO from EUTRAN Command P1. Handover Initiation P2. Handover Required P3. Forward Relocation Request P4. Relocation Request P4a. Relocation Request Response P5. Forward Reloc. Response P6b. Handover Command P7. HO from EUTRAN command Complete Map Partial Map Message Flow Current HO Preparation Signaling Proposed HO Preparation Signaling P6a. S-SGW Tunnel Setup Handover Preparation Handover Execution Figure A.16: Proposed Handover Signal mapping for LTE to 3G/2G Inter-RAT HO without Target SGW and Direct Tunnel. 233 UE S-NG-RAN T-NG-RAN S-AMF T-AMF T-UPF SeMMu S-UPF P2. Handover required P3. T-AMF Selection P4. Namf_Communication_CreateUEContext Request P1. Handover Initiation P6. UPF Selection P7a. N4 Session Establishment P8. PDU Handover Response Supervision P5. Nsmf_PDUSession_UpdateSMContext Request P7b. Nsmf_PDUSession_UpdateSMContext Response P9. Handover Request P10. Handover Request Acknowledge P11. Nsmf_PDUSession_UpdateSMContext Request P12a. N4 Session Establishment P12b. N4 Session Establishment P12c. Nsmf_PDUSession_UpdateSMContext Response P13. Namf_Communication_CreateUEContext Response Decision for N2 HO P14. Handover Command P15. Handover Command Figure A.17: Proposed Handover Signaling 5G Inter NG-RAN N2 based Handover. 234 UE E-UTRAN NG-RAN AMF T-SeMMu S-GW PGW-C + SeMMu PGW-U + UPF P2. Handover required P1. Handover Initiation P8a. Indirect Data Forwarding Tunnel P9. Nsmf_PDUSession_UpdateSMContext Request P6. Handover Request P7. Handover Request Acknowledge P11(13). HO Command P10(a). N4 Session Modification Handover Decision P3. Nsmf_PDUSession_Context Request P5. Relocation Request P8b. Relocation Response P11. N4 Session Modification P10b(12). Nsmf_PDUSession_UpdateSMContext Response P4. Nsmf_PDUSession_Context Response P12(14). HO Command Figure A.18: Proposed Signaling for 5G core to EPS Handover with N26 Interface. 235 UE E-UTRAN NG-RAN SeMMu AMF S-GW PGW-C +SeMMu PGW-U + UPF P2. Handover required P1. Handover Initiation P7(8). Handover Request P8(9). Handover Request Acknowledge P12(13)b. HO Command P5(6)a. N4 Session Modification Handover Decision P4. Nsmf_PDUSession_UpdateSMContext Request P3. Relocation Request P11(12). Relocation Response P13(14). HO Command h-PCF + hPCRF P5. PDU Session Modification P9(10). Nsmf_PDUSession_UpdateSMContext Request P5(6)b. Nsmf_PDUSession_UpdateSMContext Response P10(11)a. Nsmf_PDUSession_UpdateSMContext Response P10(11)b. N4 Session Modification P12(13)a. Indirect Data Forwarding Tunnel Creation Figure A.19: Proposed Signaling for EPS to 5G Core Handover with N26 Interface. 236 UE S-RAN T-RAN S-SeMMu T-AMF S-SGW T-SeMMu PGW-U + UPF P1. Handover cancel P7a. HO Cancel Acknowledgement Source RAN decides to cancel HO P3. N2 Release Procedure P2. Relocation Cancel Request P6. Relocation Cancel Response P4. Delete Session Request + Delete Indirect Forwarding Tunnel P7b. Delete Indirect Forwarding Tunnel P5a. Release UPF Resources P5b. Delete Session Response Figure A.20: Proposed Signaling for EPS to 5G Core Handover Cancel. 237 UE S-RAN T-RAN S-AMF T-SeMMu S-(PGW-C/SeMMu) T-SGW PGW-U + UPF P1. Handover cancel P6. HO Cancel Acknowledgement Source RAN decides to cancel HO P3. S1 Release Procedure P2. Relocation Cancel Request P4b. Relocation Cancel Response P4c. Delete Indirect Forwarding Tunnel P4a. Delete Existing Session P7. Delete Indirect Forwarding Tunnel Figure A.21: Proposed Signaling for 5G Core to EPS Handover Cancel. 238 [42] P. Fan, J. Zhao, and C.-L. I, “5G high mobility wireless communications: Challenges and solutions,” China Communications, vol. 13, no. 2, pp. 1–13, 2016. [43] S. Ferretti, V. Ghini, and F. Panzieri, “A survey on handover management in mobility architectures,” Computer Networks, 2016. [44] M. Zekri, B. Jouaber, and D. Zeghlache, “A review on mobility management and vertical handover solutions over heterogeneous wireless networks,” Computer Communications, no. 17, 2012. [45] 3GPP, “5G; System architecture for the 5G System (5GS) (3GPP TS 23.501 version 15.8.0 Release 15),” pp. 1–251, 2020. [Online]. Available: https: //portal.etsi.org/TB/ETSIDeliverableStatus.aspx [46] P. Rost, A. Banchs, I. Berberana, M. Breitbach, M. Doll, H. Droste, C. Mannweiler, M. A. Puente, K. Samdanis, and B. Sayadi, “Mobile network architecture evolution toward 5G,” IEEE Communications Magazine, no. 5, 2016. [47] I. F. Akyildiz, S. Nie, S.-C. Lin, and M. Chandrasekaran, “5G roadmap: 10 key enabling technologies,” Computer Networks, 2016. [48] S. E. Elayoubi, M. Fallgren, P. Spapis, G. Zimmermann, D. Martin-Sacristan, C. Yang, S. Jeux, P. Agyapong, L. Campoy, Y. Qi, and S. Singh, “5G service requirements and operational use cases: Analysis and METIS II vision,” EUCNC 2016 - European Conference on Networks and Communications, pp. 158–162, 2016. [49] W. Khawaja, I. Guvenc, D. W. Matolak, U.-C. Fiebig, and N. Schneckenberger, “A Survey of Air-to-Ground Propagation Channel Modeling for Unmanned Aerial Vehicles,” IEEE Communications Surveys & Tutorials, 2019. [50] B. Li, Z. Fei, and Y. Zhang, “UAV communications for 5G and beyond: Recent advances and future trends,” IEEE Internet of Things Journal, vol. 6, no. 2, pp. 2241– 2263, 2019. [51] H. Wymeersch, G. Seco-Granados, G. Destino, D. Dardari, and F. Tufvesson, “5G mmwave positioning for vehicular networks,” IEEE Wireless Communications, vol. 24, no. 6, pp. 80–86, 2017. [52] M. Jaber, M. A. Imran, R. Tafazolli, and A. Tukmanov, “5G Backhaul Challenges and Emerging Research Directions: A Survey,” IEEE Access, 2016. 245 [53] D. Liu and H. Chan, “RFC 7429 Distributed Mobility Management: Current Practices and Gap Analysis,” pp. 1–34, 2015. [54] C. Chen, Y.-T. Lin, L.-H. Yen, M.-C. Chan, and C.-C. Tseng, “Mobility management for low-latency handover in SDN-based enterprise networks,” in IEEE Wireless Communications and Networking Conference, 2016. [55] S. Wang, J. Xu, N. Zhang, and Y. Liu, “A Survey on Service Migration in Mobile Edge Computing,” IEEE Access, vol. 6, pp. 23 511–23 528, 2018. [56] T. Bai and R. W. Heath, “Coverage analysis for millimeter wave cellular networks with blockage effects,” IEEE Global Conference on Signal and Information Processing, GlobalSIP 2013 - Proceedings, pp. 727–730, 2013. [57] L. Zanzi and V. Sciancalepore, “On Guaranteeing End-to-End Network Slice Latency Constraints in 5G Networks,” Proceedings of the International Symposium on Wireless Communication Systems, pp. 1–6, 2018. [58] R. Molina-Masegosa and J. Gozalvez, “* LTE-V for Sidelink 5G V2X Vehicular Communications,” IEEE Vehicular Technology Magazine, vol. 12, no. 4, pp. 30–39, 2017. [59] E. Dahlman, S. Parkvall, J. Sköld, and P. Beming, 3G Evolution: HSPA and LTE for Mobile Broadband, 1st ed. Elsevier Ltd., 2007. [60] Ericsson, “Ericsson Mobility Report,” Tech. Rep., 2016. [Online]. Available: https://www.ericsson.com/res/docs/2016/ericsson-mobility-report-2016.pdf [61] Alcatel Lucent, “The LTE Network Architecture,” 2009. [Online]. Available: http://www.cse.unt.edu/~rdantu/FALL_2013_WIRELESS_NETWORKS/ LTE_Alcatel_White_Paper.pdf [62] E. Dahlman, S. Parkvall, and J. Sköld, 4G: LTE Advanced Pro and the Road to 5G, 3rd ed. Academic Press, 2016. [63] D. P. Ibarra Barreno, “LTE / WIFI AGGREGATION IMPLEMENTATION AND EVALUATION,” Ph.D. dissertation, UPC, 2017. [64] 3GPP, “TS 136 361 - V14.1.0 - LTE; Evolved Universal Terrestrial Radio Access (EUTRA); LTE-WLAN Radio Level Integration Using Ipsec Tunnel (LWIP) encapsulation; Protocol specification (3GPP TS 36.361 version 14.1.0 Release 14),” pp. 1–12, 2018. 246 [65] K. Samdanis, T. Taleb, and S. Schmid, “Traffic offload enhancements for eUTRAN,” IEEE Communications Surveys and Tutorials, vol. 14, no. 3, pp. 884–896, 2012. [66] C. B. Sankaran, “Data offloading techniques in 3GPP Rel-10 networks: A tutorial,” IEEE Communications Magazine, vol. 50, no. 6, pp. 46–53, 2012. [67] ITU-T, “Framework of vertical multihoming in IPv6-based next generation networks,” 2011. [68] R. Irmer, H. Droste, P. Marsch, G. P. Fettweis, S. Brueck, H.-P. Mayer, L. Thiele, and V. Jungnickel, “Coordinated multipoint: Concepts, performance, and field trial results,” Commun. Mag., no. February, pp. 102–112, 2011. [Online]. Available: http://ieeexplore.ieee.org/xpls/abs{_}all.jsp?arnumber=5706317 [69] C. Perkins, “RFC 6275 Mobility support in IPv6,” pp. 1–169, 2011. [70] R. Koodli, “RFC 4068 Fast Handovers for Mobile IPv6,” pp. 1–42, 2005. [71] H. Soliman, C. Castelluccia, K. Elmalki, and L. Bellier, “RFC 5380 HMIPv6,” pp. 1–25, 2008. [72] K. Leung, “RFC 5213 Proxy Mobile IPv6,” pp. 1–92, 2008. [73] S. Gundavelli et al., “Proxy Mobile IPv6,” RFC 5213, pp. 1–92, 2008. [74] C. Bernados, “Proxy Mobile IPv6 Extensions to Support Flow Mobility,” RFC 7864, pp. 1–19, 2016. [75] 3GPP, “Universal Mobile Telecommunications System (UMTS); LTE; Proxy Mobile IPv6 (PMIPv6) based Mobility and Tunnelling protocols; Stage 3 (3GPP TS 29.275 version 8.6.0 Release 8),” pp. 1–73, 2010. [76] H. N. Nguyen and C. Bonnet, “Scalable proxy mobile IPv6 For heterogeneous wireless networks,” Proceedings of the International Conference on Mobile Technology, Applications, and Systems, Mobility’08, 2008. [77] F. Giust, L. Cominardi, and C. Bernardos, “Distributed mobility management for future 5G networks: overview and analysis of existing approaches,” IEEE Communications Magazine, vol. 53, no. 1, pp. 142–149, 1 2015. [78] A. Ford, C. Raiciu, M. Handley, S. Barre, and J. Iyengar, “MPTCP RFC 6182,” pp. 1–28, 2011. 247 [79] A. Ford et al., “TCP Extensions for Multipath Operation with Multiple Addresses,” RFC 6824, pp. 1–64, 2013. [80] A. Ravanshid et al., “Multi-connectivity functional architectures in 5G,” in IEEE International Conference on Communications Workshops (ICC), 2016. [81] T. Klein, “Enhancements to Improve the Applicability of Multipath TCP to Wireless Access Networks,” IETF (draft), no. c, pp. 1–25, 2011. [82] S. Zannettou, M. Sirivianos, and F. Papadopoulos, “Exploiting path diversity in datacenters using MPTCP-aware SDN,” Proceedings - IEEE Symposium on Computers and Communications, pp. 539–546, 2016. [83] C. D. Phung et al., “MPTCP robustness against large-scale man-in-the-middle attacks,” Computer Networks, vol. 164, p. 106896, 2019. [Online]. Available: https://doi.org/10.1016/j.comnet.2019.106896 [84] Y. Liu, A. Neri, A. Ruggeri, and A. M. Vegni, “A MPTCP-Based Network Architecture for Intelligent Train Control and Traffic Management Operations,” IEEE Trans. Intell. Transp. Syst., vol. 18, no. 9, pp. 2290–2302, 2017. [85] P. Natarajan, F. Baker, C. Systems, P. D. Amer, and J. T. Leighton, “SCTP : What , Why , and How,” IEEE Internet Comput., vol. 13, no. 5, pp. 81–85, 2009. [86] C. Raiciu, M. Handly, and D. Wischik, “Coupled Congestion Controol for Multipath Transport Protocols,” IETF RFC6356, pp. 1–12, 2011. [87] D. Wischik et al., “Design, implementation and evaluation of congestion control for multipath TCP,” Proceedings of NSDI 2011: 8th USENIX Symposium on Networked Systems Design and Implementation, pp. 99–112, 2011. [88] P. Ignaciuk and M. Morawski, “Discrete-time MPTCP flow control for channels with diverse delays and uncertain capacity,” 2018 22nd International Conference on System Theory, Control and Computing, ICSTCC 2018 - Proceedings, pp. 722–727, 2018. [89] X. Wei, C. Xiong, and E. Lopez, “MPTCP proxy mechanisms,” Internet Engineering Task Force, Internet-Draft draft-wei-mptcp-proxy-mechanism-02, Jun. 2015, work in Progress. [Online]. Available: https://datatracker.ietf.org/doc/html/ draft-wei-mptcp-proxy-mechanism-02 [90] R. Stewart, “Stream Control Transmission Protocol,” RFC 4960, pp. 1–152, 2007. 248 [91] A. De La Oliva, A. Banchs, I. Soto, T. Melia, and A. Vidal, “An overview of IEEE 802.21: Media-independent handover services,” pp. 96–103, 2008. [92] L. Eastwood et al., “Mobility Using IEEE 802.21 in a Heterogeneous IEEE 802.16/802.11-based, IMT-Advanced (4G) Network,” IEEE Wireless Communications, no. Apr., pp. 26–34, 2008. [93] IEEE, IEEE 802.21c-2014: IEEE Standard for Local and metropolitan area networks — Part 21 : Media Independent Handover Services Amendment 3: Optimized Single Radio Handovers, 2014. [94] J.-S. Wu, S.-F. Yang, and B.-J. Hwang, “A terminal-controlled vertical handover decision scheme in IEEE 802.21-enabled heterogeneous wireless networks Jung-Shyr,” Int. J. Commun. Syst., vol. 22, pp. 819–834, 2009. [95] R. Qureshi, A. Dadej, and Q. Fu, “Issues in 802.21 mobile node controlled handovers,” in Australasian Telecommunication Networks and Applications Conference, 2007. [96] 3GPP, “3gpp TS 36.331 – 3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Evolved Universal Terrestrial Radio Access (E-UTRA); Radio Resource Control (RRC); Protocol specification (Release 10) The,” no. June, 2011. [97] D. Xenakis et al., “Mobility management for femtocells in LTE-advanced: Key aspects and survey of handover decision algorithms,” IEEE Communications Surveys and Tutorials, vol. 16, no. 1, pp. 64–91, 2014. [98] J. Zhao, “A Survey of Reconfigurable Intelligent Surfaces: Towards 6G Wireless Communication Networks with Massive MIMO 2.0,” pp. 1–7, 2019. [Online]. Available: http://arxiv.org/abs/1907.04789 [99] 3GPP, “5G; Procedures for the 5G System (5GS) (3GPP TS 23.502 version 15.8.0 Release 15),” pp. 1–362, 2020. [100] 3GPP, “5G NR; Overall description; Stage-2 (3GPP TS 38.300 version 15.8.0 Release 15),” vol. 1, pp. 1–102, 2020. [Online]. Available: https://portal.etsi.org/TB/ ETSIDeliverableStatus.aspx [101] ETSI and 3GPP, “ETSI TS 137 340 v15.5.0,” Tech. Rep., 2019. 249 [102] S. Jung and J. Kim, “A new way of extending network coverage: Relay-assisted D2D communications in 3GPP,” ICT Express, vol. 2, no. 3, pp. 117–121, 9 2016. [103] M. Kantor, R. State, T. Engel, and G. Ormazabal, “A policy-based per-flow mobility management system design,” in Proceedings of the Principles, Systems and Applications on IP Telecommunications - IPTComm ’15. New York, New York, USA: ACM Press, 2015, pp. 35–42. [Online]. Available: http://doi.acm.org/10.1145/ 2843491.2843835http://dl.acm.org/citation.cfm?doid=2843491.2843835 [104] M. Gramaglia, A. Banchs, V. Sciancalepore, Z. Yousaf, C. Mannweiler, L. Yu, B. Sayadi, M.-L. Alberi Morel, R. L. Silva, M. R. Crippa, D. V. Hugo, P. Arnold, V. Frederikos, and I. L. Pavon, “Definition of connectivity and QoE / QoS management mechanisms – 5G Norma deliverable D5.1,” 2016. [105] G. Schütz, “A k-Cover Model for Reliability-Aware Controller Placement in SoftwareDefined Networks,” in Computational Science – ICCS 2019. Springer International Publishing, 2019, pp. 604–613. [106] S. Kuklinski, Y. Li, and K. T. Dinh, “Handover management in SDN-based mobile networks,” in IEEE Globecom Workshops (GC Wkshps), 2014. [107] F. Meneses, C. Guimares, D. Corujo, and R. L. Aguiar, “SDN-based Mobility Management: Handover Performance Impact in Constrained Devices,” in 2018 9th IFIP International Conference on New Technologies, Mobility and Security (NTMS). IEEE, 2 2018, pp. 1–5. [Online]. Available: http://ieeexplore.ieee.org/document/ 8328716/ [108] T. D. Assefa et al., “SDN-based local mobility management with X2-interface in femtocell networks,” IEEE Int. Work. Comput. Aided Model. Des. Commun. Links Networks, CAMAD, pp. 3–8, 2017. [109] S. Basloom and N. Akkari, “Mobility Management in SDN and NFV-based NextGeneration Wireless Networks : An Overview and Qualitative Evaluation,” 2018 1st International Conference on Advanced Research in Engineering Sciences (ARES), pp. 1–8. [110] I. Elgendi, K. S. Munasinghe, and A. Jamalipour, “A Three-Tier SDN based distributed mobility management architecture for DenseNets,” 2016 IEEE International Conference on Communications, ICC 2016, 2016. 250 [111] Q. Li, H. Niu, A. Papathanassiou, and G. Wu, “Edge Cloud and Underlay Networks: Empowering 5G Cell-Less Wireless Architecture,” in Proceedings of European Wireless 2014, 20th European Wireless Conference, 2014. [112] ETSI, “Mobile Edge Computing ( MEC ); End to End Mobility Aspects,” Tech. Rep., 2017. [113] A. Mtibaa, R. Tourani, S. Misra, J. Burke, and L. Zhang, “Towards edge computing over named data networking,” Proceedings - 2018 IEEE International Conference on Edge Computing, EDGE 2018 - Part of the 2018 IEEE World Congress on Services, 2018. [114] P. Mach and Z. Becvar, “Mobile Edge Computing: A Survey on Architecture and Computation Offloading,” 2017. [Online]. Available: http://arxiv.org/abs/1702. 05309%0Ahttp://dx.doi.org/10.1109/COMST.2017.2682318 [115] ETSI, “3GPP TS 138 331 - V15.2.1 - 5G; NR; Radio Resource Control (RRC),” vol. 1, 2018. [116] T. Nakamura, S. Nagata, A. Benjebbour, Y. Kishiyama, T. Hai, S. Xiaodong, Y. Ning, and L. Nan, “Trends in small cell enhancements in LTE advanced,” IEEE Communications Magazine, vol. 51, no. 2, pp. 98–105, 2 2013. [117] F. A. A. Emam, M. E. Nasr, and S. E. Kishk, “Coordinated Handover Signaling and Cross-Layer Adaptation in Heterogeneous Wireless Networking,” Mob. Networks Appl., vol. 25, pp. 285–299, 2020. [118] F. A. A. Emam, M. E. Nasr, and S. E. Kishk, “Context-aware parallel handover optimization in heterogeneous wireless networks,” Ann. Telecommun., vol. 75, pp. 43–57, 2020. [119] A. Al-rubaye and J. Seitz, “A Cross-Layer Mobility Management with Multi-Criteria Decision Making,” 2016 Eighth Int. Conf. Ubiquitous Futur. Networks, pp. 821–826, 2016. [120] N. Nikaein et al., “Demo – Closer to Cloud-RAN : RAN as a Service,” ACM Mobicom, pp. 193–195, 2015. [121] A. Outtagarts et al., “When IT meets Telco : RAN as a Service,” 2015 IEEE/ACM 8th Int. Conf. Util. Cloud Comput., pp. 422–423, 2015. 251 [122] D. Sabella et al., “RAN as a Service: Challenges of Designing a Flexible RAN Architecture in a Cloud-based Heterogeneous Mobile Network,” 2013 Futur. Netw. Mob. Summit, pp. 1–8, 2013. [123] L. Liu et al., “Analysis of Handover Performance Improvement in Cloud-RAN Architecture,” 7th Int. Conf. Commun. Netw. China, pp. 850–855, 2012. [124] V. Passast et al., “Dynamic RAT Selection and Pricing for Efficient Traffic Allocation in 5G HetNets,” IEEE ICC, pp. 1–6, 2019. [125] S. Goudarzi et al., “A hybrid intelligent model for network selection in the industrial Internet of Things,” Appl. Soft Comput. J., vol. 74, pp. 529–546, 2019. [126] J. Wang, J. Weitzen, O. Bayat, V. Sevindik, and M. Li, “Interference coordination for millimeter wave communications in 5G networks for performance optimization,” Eurasip Journal on Wireless Communications and Networking, vol. 2019, no. 1, 2019. [127] D. Calabuig, S. Barmpounakis, S. Gimenez, A. Kousaridas, T. R. Lakshmana, J. Lorca, P. Lunden, Z. Ren, P. Sroka, E. Ternon, V. Venkatasubramanian, and M. Maternia, “Resource and Mobility Management in the Network Layer of 5G Cellular Ultra-Dense Networks,” IEEE Communications Magazine, vol. 55, no. 6, pp. 162–169, 2017. [128] O. N. C. Yilmaz et al., “Smart mobility management for D2D communications in 5G networks,” 2014 IEEE Wirel. Commun. Netw. Conf. Work. WCNCW 2014, pp. 219–223, 2014. [129] K. Ouali and B. Kervella, “An Efficient D2D Handover Management Scheme for SDNbased 5G networks,” 2020 IEEE 17th Annu. Consum. Commun. Netw. Conf., pp. 1–6, 2020. [130] R. Klempous and J. Nikodem, Smart Innovations in Engineering and Technology, 2020, vol. 15. [Online]. Available: http://link.springer.com/10.1007/978-3-030-32861-0 [131] S. Barua and R. Braun, “Mobility management of D2D communication for the 5G cellular network system: A study and result,” 2017 17th Int. Symp. Commun. Inf. Technol. Isc. 2017, pp. 1–6, 2017. [132] S. Barua and R. Braun, “A novel approach of mobility management for the D2D communications in 5G mobile cellular network system,” in 2016 18th Asia-Pacific Network Operations and Management Symposium (APNOMS). IEEE, 10 2016, pp. 1–4. [Online]. Available: http://ieeexplore.ieee.org/document/7737272/ 252 [133] 3GPP and ETSI, “ETSI TS 123 401,” Tech. Rep., 2015. [134] S. Oh, B. Ryu, and Y. Shin, “EPC signaling load impact over S1 and X2 handover on LTE-Advanced system,” 3rd World Congress on Information and Communication Technologies, WICT 2013, pp. 183–188, 2013. [135] D. Astely, E. Dahlman, G. Fodor, S. Parkvall, and J. Sachs, “LTE Release 12 and Beyond David,” Ieee Wirel. Commun. Mag., no. July, pp. 154–160, 2013. [136] W. Sun and J. Liu, “Coordinated multipoint-based uplink transmission in internet of things powered by energy harvesting,” IEEE Internet Things J., vol. 5, no. 4, pp. 2585–2595, 2018. [137] J. Lee, Y. Kim, H. Lee, B. Ng, D. Mazzarese, J. Liu, W. Xiao, and Y. Zhou, “Coordinated multipoint transmission and reception in LTE-advanced systems,” IEEE Commun. Mag., vol. 50, no. 11, pp. 44–50, 2012. [138] C. Shen and M. Van Der Schaar, “A learning approach to frequent handover mitigations in 3GPP mobility protocols,” IEEE Wirel. Commun. Netw. Conf. WCNC, pp. 1–6, 2017. [139] A. Ahmed, L. M. Boulahia, and D. Gaïti, “Enabling vertical handover decisions in heterogeneous wireless networks: A state-of-the-art and a classification,” IEEE Commun. Surv. Tutorials, vol. 16, no. 2, pp. 776–811, 2014. [140] A. Santoyo-González and C. Cervelló-Pastor, “Latency-aware cost optimization of the service infrastructure placement in 5G networks,” Journal of Network and Computer Applications, vol. 114, no. February, pp. 29–37, 2018. [Online]. Available: https://doi.org/10.1016/j.jnca.2018.04.007 [141] I. Leyva-Pupo, A. Santoyo-González, and C. Cervelló-Pastor, “A framework for the joint placement of edge service infrastructure and user plane functions for 5G,” Sensors (Switzerland), vol. 19, no. 18, 2019. [142] R. Alkhansa, H. Artail, and D. M. Gutierrez-Estevez, “LTE-WiFi carrier aggregation for future 5G systems: A feasibility study and research challenges,” Procedia Computer Science, vol. 34, pp. 133–140, 2014. [Online]. Available: http://dx.doi.org/10.1016/j.procs.2014.07.068 [143] M. A. Ferrag, L. Maglaras, A. Argyriou, D. Kosmanos, and H. Janicke, “Security for 4G and 5G Cellular Networks: A Survey of Existing Authentication 253 and Privacy-preserving Schemes,” pp. 1–24, 2017. [Online]. Available: http: //arxiv.org/abs/1708.04027 [144] M. Jawad Alam and M. Ma, “DC and CoMP Authentication in LTE-Advanced 5G HetNet,” IEEE Global Communications Conference, GLOBECOM 2017 - Proceedings, pp. 1–6, 2018. [145] G. Qiao, S. Leng, K. Zhang, and K. Yang, “Joint Deployment and Mobility Management of Energy Harvesting Small Cells in Heterogeneous Networks,” IEEE Access, vol. 5, pp. 183–196, 2017. [146] A. Habbal, S. Goudar, and S. Hassan, “Context-aware Radio Access Technology Selection Approach in 5G Ultra Dense Networks,” IEEE Access, 2017. [147] 3GPP, “TS22.261: Service requirements for the 5G system (Stage 1),” 2018. [148] A. Sadeghian, L. Sundaram, D. Z. Wang, W. F. Hamilton, K. Branting, and C. Pfeifer, “Semantic Edge Labeling over Legal Citation Graphs,” in LTDCA, 2018. [149] M. Boban, A. Kousaridas, K. Manolakis, J. Eichinger, and W. Xu, “Use Cases, Requirements, and Design Considerations for 5G V2X,” pp. 1–10, 2017. [Online]. Available: http://arxiv.org/abs/1712.01754 [150] G. Report, “Study on MEC Support for V2X Use Cases,” vol. 1, pp. 1–19, 2018. [151] G. A. Zhang, J. Y. Gu, Z. H. Bao, C. Xu, and S. B. Zhang, “Efficient Signal Detection for Cognitive Radio Relay Networks Under Imperfect Channel Estimation,” European Transactions on Telecommunications, vol. 25, no. 3, pp. 294–307, 2014. [152] D. Raychaudhuri, K. Nagaraja, N. Brunswick, and A. Venkataramani, “MobilityFirst : A Robust and Trustworthy MobilityCentric Architecture for the Future Internet,” ACM SIGMobile Mobile Computing and Communication Review (MC2R), pp. 1–12, 2012. [Online]. Available: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1. 440.2259&rep=rep1&type=pdf [153] K. Pentikousis, Y. Wang, and W. Hu, “Mobileflow: Toward software-defined mobile networks,” IEEE Communications Magazine, vol. 51, no. 7, pp. 44–53, 2013. [154] P. Marsch et al., “5G RAN Architecture and Functional Design,” 2016. [Online]. Available: https://metis-ii.5g-ppp.eu/documents/white-papers/ 254 Networks-With a Focus on Propagation Models,” IEEE Transactions on Antennas and Propagation, vol. 65, no. 12, pp. 6213–6230, 2017. [218] Keysight, “Understanding the 5G NR Physical Layer,” Tech. Rep., 2017. [Online]. Available: https://www.keysight.com/upload/cmc_upload/All/Understanding_the_ 5G_NR_Physical_Layer.pdf [219] S. Sun, T. S. Rappaport, M. Shafi, P. Tang, J. Zhang, and P. J. Smith, “Propagation Models and Performance Evaluation for 5G Millimeter-Wave Bands,” IEEE Transactions on Vehicular Technology, vol. 67, no. 9, pp. 8422–8439, 2018. [220] 3GPP, “TR 138 901 - V14.0.0 - 5G; Study on channel model for frequencies from 0.5 to 100 GHz (3GPP TR 38.901 version 14.0.0 Release 14),” Tech. Rep., 2017. [Online]. Available: http://www.etsi.org/standards-search [221] T. Report, “TR 138 901 - V15.0.0 - 5G; Study on channel model for frequencies from 0.5 to 100 GHz (3GPP TR 38.901 version 15.0.0 Release 15),” vol. 0, 2018. [222] R. Jain, D. Chiu, and W. Hawe, “A Quantitative Measure Of Fairness And Discrimination For Resource Allocation In Shared Computer Systems,” 1998. [Online]. Available: http://arxiv.org/abs/cs/9809099 [223] C. Kim, R. Ford, and S. Rangan, “Joint interference and user association optimization in cellular wireless networks,” Conference Record - Asilomar Conference on Signals, Systems and Computers, vol. 2015-April, pp. 511–515, 2015. [224] Y. Liu, M. Derakhshani, and S. Lambotharan, “Dual Connectivity in Backhaul-limited Massive-MIMO HetNets : User Association and Power Allocation.” [225] G. Pocovi, S. Barcos, H. Wang, K. I. Pedersen, and C. Rosa, “Analysis of Heterogeneous Networks with Dual Connectivity in a Realistic Urban Deployment,” no. May 2016, 2015. [226] M. Eisen and A. Ribeiro, “Optimal Wireless Resource Allocation with Random Edge Graph Neural Networks,” pp. 1–15, 2019. [Online]. Available: http: //arxiv.org/abs/1909.01865 261