Energy-efficient Optical Aggregation Enabled by Digital Sub-Carriers

979-8-3315-9777-1/25/$31.00 ©2025 IEEE Energy-efficient Optical Aggregation Enabled by Digital Sub-Carriers P. Castoldi, N. Sambo Scuola Superiore Sant’Anna Pisa, Italy [email protected] M. M. Hosseini, A. Napoli Nokia, Optical Networks Germany J. Pedro, N. Costa Nokia, Optical Networks Portugal M. Quagliotti TIM Torino, Italy E. Riccardi Fibercop Torino, Italy J. A. Hern´ andez, D. de la Osa Mostazo, ´ O. Gonz´ alez de Dios Transport Networks, Telef´ onica Innovaci´ on Digital Madrid, Spain Abstract—In this work, we leverage all-optical traffic aggregation enabled by digital sub-carrier multiplexing, eliminating electronic aggregation interfaces. The quality of transmission estimation shows a minimal impact on transmission performance while significantly reducing energy consumption. Index Terms—Energy consumption, traffic aggregation, alloptical, point-to-multi-point, P2MP, digital sub-carrier multiplexing. I. INTRODUCTION Nowadays the ICT sector is accounting for around 3% of global greenhouse emissions and 5−9% of electricity use [1]. ICT energy consumption is expected to increase given the traffic growth. The European Commission estimates that the ICT footprint could increase up to 14% of global emissions by 2040. Thus, solutions to reduce power consumption of world network infrastructures are fundamental. In [2], energy consumption of telecommunications networks is discussed and several solutions to realize more energy-efficient networks have been presented. A large contribution to the energy requirements is caused by router interfaces and by operations performed in the electronic domain (e.g., traffic aggregation). For instance, different network segments – such as access and regional – are interconnected by routers or electronic switches. At an edge node, lower-rate traffic flows are aggregated electronically and then routed in the metro-regional or national segment. Performing traffic aggregation in the optical domain may strongly reduce the energy requirements at the edge nodes. In this paper, we analyze digital sub-carrier multiplexing (DSCM) to all-optically aggregate traffic at the edge node of a metro-regional/national segment to avoid energy-hungry electronic (router) interfaces. This requires that optical flows transparently i) traverse access segments, ii) are aggregated through a passive optical coupler, iii) then, traverse the metroregional/national segment. The Quality of Transmission (QoT) of such flows is here estimated, resulting in a negligible impact due to the propagation in the access segment and due to the optical traffic aggregation, while significant benefits are achieved in terms of energy and cost savings. II. OPTICAL TRAFFIC AGGREGATION WITH DSCM DSCM [3] has been recently proposed as a transmission technique that offers an additional layer of multiplexing: i.e., the digital sub-carriers (DSCs). One of the key advantages of DSCM is the possibility of optically aggregating traffic. This may pave the way to relevant changes in traffic aggregation as illustrated in Fig. 1a. Here, several transceivers are associated with different access segment traffic collection points, e.g., serving traffic from optical line terminals (OLTs) collecting traffic from several passive optical networks (PONs) for fixed access, a baseband unit (BBU) or a cell site router for mobile radio site. The gathered traffic is allocated to a DSC or a subset of available DSCs. The signals flow over the access segment and then enter the passive coupler where the DSCs are aggregated all-optically. An Erbium-doped fiber amplifier (EDFA) might be needed to compensate for the loss due to the propagation from the access to the edge and for the losses due to the coupler. Finally, the aggregated signal flows within the metro-regional or national network. Such an architecture can be referred to as point-to-multipoint (P2MP) and it contrasts with the point-to-point (P2P) architecture reported in Fig. 1b. Here, with P2P, each traffic flow from access is received by a dedicated transceiver at the edge; then, a router aggregates the traffic, and another (higherrate) transceiver (maybe even installed into a transponder) generates the signal that is routed over the metro-regional/national segment. The comparison of the two architectures in Fig. 1 shows that optical aggregation may strongly reduce power consumption because i) transceiver interfaces can be saved at the edge since signals are by-passed and not terminated; ii) router interfaces are saved at the edge since traffic is aggregated alloptically. In the case of optical aggregation, QoT estimation should account for i) signal propagation within the access segments, ii) the losses due to the passive coupler, and iii) the propagation within the metro-regional/national segment. DSCs originating in different access nodes are aggregated in a single DSCM signal and terminated at a single destination metroregional/national node, where a router will distribute the subtraffic flows to other segments. Note that as the DSCs – before Fig. 1: (a) Optical and (b) Electronic aggregation approaches. Fig. 2: Considered network scenario. being aggregated – may follow different paths, they may also experience different attenuation. However, as long as the optical power imbalance among all received DSCs is within 10 dB, each individual SC can be correctly detected [4]. It is worth noticing that an optical power equalization strategy can be adopted to keep the power imbalance within the transceiver requirements. In the next section, QoT estimation results are presented for a specific network scenario. III. RESULTS FOR QOTESTIMATION The network scenario of Fig. 2 depicts a metroregional/national topology with links’ (and spans’) length sourced from the EU SEASON project [5]. DSCM is assumed with 16 DSCs for a maximum overall capacity of 400G per transceiver. Each metro-regional/national node serves four fixed or mobile access segment collection points with links of 10 km, allowing a maximum capacity of 100G per 10 km link. DSCs are generated in the access segments, aggregated all-optically in a metro-regional/national node, and then terminated to another metro-regional/national node, as in the illustrative example (in orange) in Fig. 2. The QoT estimation is based on optical-signal-to-noise ratio (OSNR) computation in a linear regime. OSNR is computed per path (e.g., assuming the worst wavelength channel). A fiber attenuation of 0.2 dB/km is considered; DSCs aggregation (i.e., at the coupler) is assumed to introduce a loss of 6 dB; metroregional/national node architecture is switch&select and each node introduces 10 dB loss (5 dB per wavelength selective switch). In the edge node, an EDFA compensates for fiber attenuation within the access and for the coupler loss. Launch power is set to -1.0 dBm. Then, an EDFA is placed after each link or fiber span compensating for fiber attenuation and node loss. OSNR requirements are retrieved from the XR Optics specifications [6] and are 24 dB for 400G DP16QAM, 20 dB for 300G DP-8QAM, and 15 dB for 200G DP-QPSK. A variable amount of operational margins [7], ranging from 0 dB to 5 dB, are considered to account for other effects such as aging and QoT model inaccuracies. The OSNR is computed for all routes starting from an access segment and terminating at a metro-regional/national node as in the example of Fig. 2. As a benchmark, the QoT in the scenario of electronic aggregation of traffic is estimated. In this case, the OSNR is computed for all routes starting in a metro-regional/national node and terminating in different metro-regional/national nodes. Fig. 3 reports the percentage of routes supporting channels using the most spectral-efficient modulation format (DP16QAM, DP-8QAM, or DP-QPSK), as a function of the operational margins. The analysis of Fig. 3 shows that using P2MP (optical aggregation) or P2P (electronic aggregation) lead to similar results. Thus, having an optical by-pass (which includes the propagation from the access, the coupler loss for traffic aggregation, and the related amplification noise) has only a minimal impact on the QoT, so that the same most spectral efficient modulation format can be selected with P2MP or P2P. Focusing on 400G DP-16QAM, around 80% of routes support it when considering no operational margin (in both P2MP and P2P approaches). However, the percentage of routes supporting 400G DP-16QAM decreases as the margins 0 20 40 60 80 100 012345 % Margins [dB] P2P P2MP 0 20 40 60 80 100 012345 % Margins [dB] P2P P2MP 0 20 40 60 80 100 012345 % Margins [dB] P2P P2MP 400G DP-16QAM 300G DP-8QAM 200G DP-QPSK Fig. 3: Percentage of routes where the most spectral-efficient modulation format is DP-16QAM, DP-8QAM, or DP-QPSK. 0 10 20 30 40 50 60 70 80 90 100 230 232 234 236 238 Traffic [Gb/s] Power Consumption [W] ∼6.9W Fig. 4: Power consumption [W] vs. traffic [Gb/s]. increase since, in this case, some routes cannot reach the OSNR requirement of 24 dB. Consequently, 300G DP-8QAM should be transmitted instead. This effect is particularly clear when increasing margins up to 3 dB. If even higher operational margins are considered, some routes that were previously feasible with 8QAM modulation format become infeasible (thus, the percentage of 8QAM routes may decrease for specific operational margins) and the use of QPSK modulation format rapidly becomes necessary. Since P2MP leads to slightly worse QoT than P2P only, the difference between the two is visible for a few cases of operational margins (1.5 dB, 4 dB, and 5 dB) only. Moreover, the resulting difference in the percentage of routes supporting 400G DP-16QAM is very small (2%). IV. ENERGY AND COST SAVINGS To evaluate the energy saving benefits of the proposed approach, i.e., optical aggregation with P2MP, we analyze the energy consumption of a router (handling up to 10 ×400G QSFP-DD service ports). In particular, power consumption measurements of an IPoWDM device are presented. The measurements were done every 15 seconds through command line interface. Note that the power consumption of the equipment without traffic and without any connected pluggable is ∼226 W. Then, pluggables are connected: when a single 100G QSFP28 pluggable is connected, the energy consumption increases by ∼4 W whereas adding one 400G CFP2 pluggable increases the energy consumption by ∼22 W. Then, tests have been done to assess energy consumption depending on traffic when the equipment is configured with 1×100G pluggable connected. Traffic has been increased by 10 Gb/s every 5 minutes. The energy consumption is reported in Fig. 4. As can be seen, by adopting optical aggregation and avoiding some of the edge routers, savings of >230W can be achieved at each aggregator node. The impact is significant considering that, for example, in Italy there are ∼3000 aggregator nodes [5], each equipped with one or more routers. This is also reflected in cost savings: assuming 0.2C/kWh as the cost of energy and the (conservative) router power consumption of the idle case, an annual operational expense savings of 1.2MC can be achieved. V. CONCLUSIONS Traffic aggregation in the optical domain – enabled by digital sub-carrier multiplexing – is here adopted to increase energy efficiency. The resulting network architecture has only a minimal impact on transmission performance while the energy savings achieved by removing routers are relevant, e.g. >230 W per router. For the Italian network, cost savings of ∼1MC per year can be achieved. ACKNOWLEDGEMENTS This work was supported by the EU under grant agreement 101096120 (SEASON) and under the Italian National Recovery and Resilience Plan (PNRR) of NextGenerationEU partnership on “Telecommunications of the Future” (PE00000001 - program ”RESTART”). REFERENCES [1] European Commission, “Supporting the green transition,” February 2020. [2] N. Sambo and et al., “Solutions to increase energy efficiency of optical networks,” in Proc. of OFC, 2024. [3] D. Welch and et al., “Point-to-multipoint optical networks using coherent digital subcarriers,” JLT, 2021. [4] G. Simon and et al., “200 Gb/s coherent point-to-multipoint coexistence with 50G-PON for next-generation optical access,” IEEE PTL, 2024. [5] D2.1, EU Project SEASON. (2024) Definition of Use Cases, Requirements and Reference Network Architecture. [6] Open XR forum Optics. (2024) Open XR Optics Transceiver Optical and Client Interface Specification. [7] K. Christodoulopoulos and et al., “Toward efficient, reliable, and autonomous optical networks: the orchestra solution [invited],” IEEE/Optica JOCN, vol. 11, no. 9, pp. C10–C24, 2019.