A Survey on Distributed Database Systems in the Era of Big Data

Engineering and Technology Journal e-ISSN: 2456-3358 Volume 10 Issue 11 November-2025, Page No.-7710-7721 DOI: 10.47191/etj/v10i11.03, I.F. – 8.482 © 2025, ETJ 7710 ETJ Volume 10 Issue 11 November 2025, 1 Kazheen Ismael Hasan A Survey on Distributed Database Systems in the Era of Big Data Kazheen Ismael Hasan1, Hajar Maseeh Yasin2 1Akre University for Applied Sciences Technical College of Informatics-Akre Information Technology Department 2Akre University for Applied Sciences, Technical College of Informatics, Department of Information Technology, Duhok,KRG - Iraq ABSTRACT: Distributed database systems have evolved to satisfy the needs of scalability, performance, and fault tolerance due to the current digital era's fast data expansion. The design concepts, benefits, and drawbacks of modern distributed database architectures—such as cloud-native systems, NoSQL, and NewSQL—are thoroughly examined in this study. With an emphasis on using artificial intelligence and machine learning approaches to improve query speed and anomaly detection, key difficulties such as data integrity, latency optimization, and safe multi-cloud integration are covered. Despite notable progress, important concerns about data privacy and synchronization in diverse settings remain, and moral leadership endures. To create more robust and accountable database systems, this study promotes a well-rounded strategy that addresses the ethical and social aspects of distributed data management and increases technical efficiency. KEYWORDS: Distributed Database Systems, Big Data Management, NoSQL and NewSQL, Query Optimization. 1. INTRODUCTION An unparalleled surge of data characterizes the digital era[1]. In 2020, the global data volume exceeded 59 zettabytes and is anticipated to attain 175 zettabytes by 2025[2], a magnitude that undermines the principles of conventional database management[3].This influx has initiated a new epoch for database systems, essential for efficiently storing, managing[4], and deriving value from extensive datasets that support contemporary analytics and decision-making[5]. Traditional relational database systems are inadequately prepared to manage the volume and diversity of big data, frequently encountering difficulties in scaling and accommodating varied[6], rapidly evolving datasets[7] In response, contemporary data architectures, ranging from distributed NoSQL databases to cloud-native database services, have emerged to offer the scalability and flexibility required for big data management [8] Nevertheless, despite these advancements, contemporary research emphasizes enduring challenges[9], including data security, regulatory compliance, and latency[10] highlighting the continued necessity for significant progress in the domain[11]. 2. BACKGROUND THEORY A dramatic increase in data volume, velocity, and variety marks the Big Data era[12]. The global datasphere is anticipated to exceed 100 zettabytes by 2025 [13], presenting considerable challenges in storing, retrieving, and managing heterogeneous data [14] Conventional relational database management systems (RDBMS) are progressively insufficient in managing the scale and variety of contemporary datasets [15] Consequently, novel paradigms such as NoSQL, NewSQL, data lakes, and distributed databases have arisen[16] These advancements signify technical innovation and a fundamental transformation in the organization, accessibility, and data optimization for realtime decision-making [17]. NoSQL databases have become prominent due to their schema flexibility, horizontal scalability, and capability to handle unstructured and semi-structured data [18] They have become indispensable in web-scale applications such as social media and IoT systems. They sacrifice ACID compliance for eventual consistency within the BASE model, potentially complicating transaction integrity [19] For instance, consistency models in NoSQL databases such as Cassandra impose performance penalties under stringent conditions [20] The conflict between scalability and consistency[21] illustrates the overarching difficulty of creating database systems that are both efficient and dependable in the context of Big Data demands [22]. NewSQL systems have emerged to bridge this gap, providing ACID transactions and SQL interfaces while ensuring the distributed scalability characteristic of NoSQL[23] Systems like Google Spanner and Cockroach DB utilize advancements such as distributed consensus and sharding to ensure global consistency[24]. Performance evaluations demonstrate that NewSQL can attain throughput similar to NoSQL while maintaining transactional integrity, rendering them suitable for high-volume, mission-critical workloads [25]. Nonetheless, challenges such as distributed query optimization and conflict resolution persist, underscoring the ongoing necessity for research to scale relational models while maintaining integrity[26] “A Survey on Distributed Database Systems in the Era of Big Data” 7711 ETJ Volume 10 Issue 11 November 2025, 1 Kazheen Ismael Hasan The expansion of diverse data types has catalyzed the emergence of data lakes and bakehouses[27]. Data lakes provide economical storage for unstructured data in its original state, facilitating large-scale analytics and machine learning applications[28].However, initial implementations exposed deficiencies in governance and query performance. The lakehouse model resolves these challenges by incorporating ACID transactions and schema management atop data lakes [28]These hybrid systems combine the adaptability of lakes with the organization of warehouses, enhancing analytical capabilities across various industries [29] Furthermore, multi-model and enable organizations to manage varied data within or across systems [30] facilitating integrated queries across relational, graph, and documentoriented sources[31]. In addition to storage, performance optimization has emerged as a paramount issue. Contemporary methodologies[32], including in-memory processing, adaptive indexing, and parallel query execution, have become prevalent. AI and machine learning are progressively incorporated into database management systems to automate tuning, optimize queries, and potentially supplant conventional components such as indexes [33] Machinelearned models have surpassed heuristic methods in query optimization and workload forecasting [34] Autonomous databases now employ AI for self-configuration, adaptation to fluctuating workloads[35], and performance optimization with minimal human intervention . This transition signifies a wider industry trend towards autonomous systems that diminish operational complexity while enhancing responsiveness and reliability[36]. Cloud-native deployments and hybrid data architectures facilitate access to advanced database systems. Cloud platforms provide flexible, scalable services that facilitate global applications without substantial infrastructure expenditure Simultaneously[37], various sectorsfrom healthcare to finance are adopting analogous data strategies, prioritizing integrated storage, real-time analytics, and adherence to data governance regulations [38].These changes signify a fundamental transformation in data infrastructure, transitioning from monolithic systems to intelligent, integrated ecosystems[39]. 3. LITERATURE REVIEW This section delineates several prior studies relevant to this review article. Therefore, the current review incorporated findings from several earlier studies to interpret the key results and proposals, enhancing the background theory. Consequently, the previous studies will be delineated chronologically from the oldest to the most recent studies, as follows: Topcu and Rmis (2020) [40]assessed the efficacy of the Riak KV NoSQL database within a distributed cluster setting utilizing the Basho-bench benchmarking instrument. They simulated diverse workloads that were read-only, updateintensive, and mixed across varying data sizes and thread counts to evaluate throughput and latency. The findings indicated that read-only operations consistently attained superior throughput and reduced latency, whereas update operations diminished performance, particularly with larger datasets. Augmenting the thread count enhanced performance; however, scalability was constrained beyond a specific data volume. Their research offers insights into enhancing Riak KV for big data applications, especially for read-intensive scenarios. Mosharraf and Adnan (2020)[41] introduced two optimization strategies for distributed Big Data systems utilising Cuckoo Filters instead of Bloom Filters. The initial scheme improves lookup efficiency post-data deletion by facilitating key removal from filters, thus addressing a significant constraint in Bloom-based systems employing eventual consistency. The second scheme enhances remote query efficiency by implementing node filters that prevent superfluous network roundtrips when remote nodes do not possess the requested data. Both methodologies were executed and evaluated on Apache Cassandra, utilizing an authentic dataset, attaining a performance enhancement of up to 100% (2x) in the circumstances involving deleted or absent data. The experiments further validated that these enhancements impose negligible CPU and network overhead, rendering the approach feasible for real-world implementation. Dioulasso and Tiendrebeogo (2020) [42]proposed a distributed Big Data storage system utilizing Distributed Hash Tables (DHT) to address the scalability and faulttolerance limitations inherent in conventional MapReduce frameworks. Their model incorporates hyperbolic geometry and Poincaré disk-based addressing to facilitate decentralized, topology-independent routing and autonomous node organization. The system implements several controller nodes via virtual addresses, improving parallel processing and load distribution while preventing single points of failure. The proposed model demonstrates enhanced robustness and scalability compared to current DHT-based architectures like ChordReduce and P2P-MapReduce without depending on inflexible network topologies. Future endeavours involve the implementation of a hybrid DHT–MapReduce framework to assess performance in practical applications. Aswal (2020) [43]examined the function of distributed database systems (DDBS) in managing extensive data across multiple sectors, including healthcare, e-commerce, and IoT. The research emphasizes the benefits of DDBS, such as scalability, real-time analytics, fault tolerance, and security. It also delineates significant challenges, including data consistency, replication, synchronization, and system complexity. A comparative analysis of systems such as Cassandra, DynamoDB, and Spanner highlights the tradeoffs among performance, cost, and deployment flexibility. Jowan et al. (2021) [44]examined the shift from conventional RDBMS to NoSQL databases prompted by the difficulties “A Survey on Distributed Database Systems in the Era of Big Data” 7712 ETJ Volume 10 Issue 11 November 2025, 1 Kazheen Ismael Hasan associated with Big Data, unstructured data, and cloud-based applications. The paper classifies NoSQL systems into four categories: key-value, document, column-family, and graph databases, each engineered for flexibility and scalability. It elucidates how NoSQL utilizes the CAP theorem and BASE principles to guarantee high availability and partition tolerance in distributed systems. The study concludes that NoSQL databases are crucial for managing contemporary application requirements that involve real-time, large-scale, and diverse data. Jinadu et al. (2021) [45]introduced a distributed database optimization model utilizing a Distributed Storage Pool (DSP) enhanced by virtualization and hybrid RAID technology to enhance service delivery in mobile and cloud Big Data applications. Their architecture utilizes semi-join operations, storage replication, and mobility transparency to improve data access efficiency and reduce latency in distributed transactions. Simulations utilizing M-TCP in WLAN environments exhibited substantial enhancements in throughput and response time relative to traditional TCP configurations. The research validates that DSP utilizing virtualization diminishes overhead and guarantees high availability and fault tolerance, rendering it appropriate for real-time, latency-sensitive cloud services. Hongwei and Ligetu (2021)[46] examined distributed storage technologies to tackle the increasing difficulties of managing big data in cloud computing settings. They highlighted the constraints of centralized systems and advocated for the implementation of object-based distributed storage and virtualization to enhance scalability, efficiency, and data security. Their system accommodates diverse data types and provides adaptable, economical, and resilient storage appropriate for rapid data expansion. The research underscores the significance of adaptive storage architecture in fulfilling the performance requirements of big data applications. Chang and Cui (2021)[47] introduced a distributed storage strategy to manage economic big data distinguished by spatial, temporal, and semantic diversity. A multilevel partitioning algorithm that integrates Geohash and Hilbert curves was introduced to enhance storage efficiency and facilitate cross-modal analysis. Their system was deployed on a NoSQL database (Cassandra) and evaluated with simulated workloads to verify resource efficiency and adherence to SLA requirements. The findings validated that their spatiotemporal-semantic-aware storage strategy markedly improves performance and adaptability for extensive economic data applications. Thamar (2023)[48] investigated how distributed computing augments big data engineering by optimizing data ingestion, processing, and analysis within contemporary data architectures. The paper examines essential distributed models MapReduce, MPP, BSP, and in-memory computing, emphasizing their advantages in scalability, velocity, and real-time processing. It examines integrating distributed principles using tools such as Apache Kafka, Spark, Delta Lake, and dbt to construct reliable and fault-tolerant pipelines. The research highlights that distributed systems are essential for handling large, rapidly evolving data in contemporary analytical settings. Zhang et al. (2024)[49] introduced MultiLog, a multivariate log-based anomaly detection technique for distributed databases. The initial extensive dataset comprising 900 million log entries encompasses 11 categories of anomalies across various nodes. MultiLog extracts sequential, quantitative, and semantic features from distributed logs, employing an LSTM with self-attention and a cluster classifier for precise detection. The methodology attained an F1 score of up to 12%, which was superior to leading techniques and diminished false positives in multi-node settings. Olusegun et al. (2024)[50] investigated the Secure MultiParty Computation (SMPC) application in cloud-based big data analytics to safeguard data privacy during collaborative processing. The research analyzed protocols, including secret sharing, homomorphic encryption, and federated learning, demonstrating their capacity to facilitate secure computations while preserving the confidentiality of individual data inputs. SMPC was utilized in practical applications such as secure machine learning and privacy-preserving queries in multicloud environments. The authors determined that although scalability and communication overhead persist as challenges, SMPC is a viable method for secure data collaboration in sensitive areas. Munawar et al. (2024)[51] systematically reviewed big data applications in smart real estate and disaster management, examining 139 studies published between 2010 and 2020. The document underscored the significance of the seven Vs volume, velocity, variety, value, veracity, variability, and visualization as essential facets of big data. It proposed cohesive frameworks demonstrating how big data can improve decision-making, service delivery, and emergency response utilizing IoT, AI, and social media analytics. The research identified persistent challenges, including data quality, system integration, and scalability in real-time, multisource contexts. Ibrahim (2024)[52] proposed a data synchronization method for heterogeneous distributed databases that amalgamate both row-oriented and column-oriented storage systems. The research presents a bi-directional synchronization model utilizing custom "Dsync" logs to monitor updates, deletions, and insertions across various databases independent of timestamps. It utilizes parallel processing, routing options, and centralized coordination to guarantee consistent, realtime data exchange among independent database environments. The methodology tackles critical issues, including format discrepancies, communication lags, and system autonomy, offering a scalable resolution for practical distributed applications. “A Survey on Distributed Database Systems in the Era of Big Data” 7713 ETJ Volume 10 Issue 11 November 2025, 1 Kazheen Ismael Hasan Gadde (2024)[53] proposed an AI-driven framework to enhance transactional integrity in distributed database systems by incorporating machine learning and anomaly detection with conventional consensus protocols. The system comprises a predictive analytics engine and an integrity monitoring unit that forecasts transaction conflicts and identifies irregularities in real-time. Experimental findings in a simulated cloud environment demonstrated a 50% enhancement in transaction throughput, a 40% decrease in response time, and an 80% reduction in integrity violations. The research validates that AI can substantially improve efficacy and dependability in distributed database systems. Aryan et al. (2024)[54] introduced a Rust-based framework for implementing Decentralized Autonomous Database Systems (DADBS) to tackle scalability and autonomy in distributed settings. The system incorporates a Proof of Work consensus mechanism, smart contracts, and a peer-to-peer network developed utilizing Rust's concurrency model and SQLite for data storage. Performance testing demonstrated a throughput of 3,000 transactions per second, high consistency, and nearly linear scalability up to 500 nodes. The findings underscore Rust's appropriateness for developing secure, efficient, decentralized databases with autonomous functionalities. Zhu et al. (2025)[55] presented RAPO, an automated optimization instrument for Redis clusters employed in distributed metadata storage systems. The tool optimizes performance by balancing loads among primary nodes via greedy and random iterative algorithms, resulting in a load distribution improvement of up to 29.36%. It also employs read-write separation strategies, such as smooth weighted round-robin, to diminish metadata read latency by as much as 30.75% during periods of high concurrency. The research validates that RAPO markedly enhances the efficiency and scalability of Redis clusters in extensive distributed settings. Sato (2025)[56] examined the evolution of database architects' roles in distributed systems, transitioning from centralized schema design to overseeing scalability, partitioning, and consistency trade-offs. The study examines fundamental patterns, including sharding, schema versioning, replication models, and the ramifications of the CAP theorem. It underscores utilizing AI-assisted tools, polyglot persistence, and cloud-native technologies for managing contemporary distributed workloads. The paper asserts architects must reconcile technical complexity with strategic design in globally distributed infrastructures. Kirino (2025) [57]analyzed the impact of distributed, cloudnative, and real-time data systems on the responsibilities of contemporary database architects. The research emphasizes essential design principles, including scalability, high availability, partitioning, and consistency models, while tackling operational challenges such as observability and fault tolerance. It delineates developing responsibilities, encompassing the management of polyglot persistence, CI/CD integration, and ethical data governance within intricate infrastructures. The paper asserts that architects currently function as strategic system designers, harmonizing performance, compliance, and interdisciplinary collaboration in distributed data environments. Evelyn (2025)[58] established an adaptive query optimization framework for heterogeneous big data environments, tackling schema diversity and system variability issues. The methodology incorporates metadata abstraction, machine learning-driven cost modelling, and federated execution planning to enhance performance across various platforms. Testing on systems such as PostgreSQL, MongoDB, Hive, and Elasticsearch demonstrated execution times up to 40% faster than conventional optimizers. The research illustrates that integrating learning algorithms with dynamic metadata improves query efficiency in intricate, distributed data systems. Ailamaki (2025) [59]investigated parallel and distributed query execution as a fundamental approach for managing extensive big data workloads. The research examined systems such as Apache Spark, Hive, Presto, and Dask, emphasizing query planning, partitioning, fault tolerance, and load balancing. Experiments demonstrated that distributed execution markedly enhanced query performance; however, data skew and network bottlenecks constrained scalability beyond a specific threshold. The document underscores the necessity for adaptive, self-optimizing architectures to maintain efficient analytics in intricate distributed settings. Adeleke (2025)[60] evaluated the efficacy of approximation algorithms for processing big data in distributed database systems. The research assessed sampling, sketching, and hybrid methodologies utilizing Apache Spark and HDFS to analyze performance, accuracy, and scalability. The findings indicated that sketch-based methodologies, such as CountMin Sketch and HyperLogLog, yielded rapid, memoryefficient estimations with minimal communication overhead. The results validate that approximation methods can enhance query efficiency while preserving acceptable error thresholds in distributed settings. Abiteboul et al. (2025) [61]examined essential optimization methodologies for distributed query execution in large-scale data systems, encompassing cost-based approaches, adaptive processing, indexing, and parallel execution. Their research demonstrated that these techniques enhance performance by minimizing latency, distributing load, and optimizing resource utilization. The authors investigated nascent methodologies such as machine learning optimization and quantum computing, which exhibit potential yet remain experimental. The study concludes that hybrid, intelligent optimization frameworks are crucial for effective and scalable distributed query execution. “A Survey on Distributed Database Systems in the Era of Big Data” 7714 ETJ Volume 10 Issue 11 November 2025, 1 Kazheen Ismael Hasan Author (Year) Focus Area Techniques Key Findings Advantages Dataset Performance Limitations Topcu & Rmis (2020)[40] Riak KV benchmarking Read/update workloads Read fast, updates slow Threads help scalability Riak, Bashobench Latency, throughput Update slow Mosharraf & Adnan (2020)[41] Filter optimization Cuckoo, Cassandra Query gain 100% Low overhead Cassandra Query time Scale issues Dioulasso et al. (2020)[42] DHT storage Virtual DHT Balanced access No single point Simulated DHT Access speed Needs real use Aswal (2020[43]) DDBS SMPC, FL Real-time support Scalable General DDBS Scalability Deployment gaps Jowan et al. (2021)[44] NoSQL shift CAP, NoSQL types Handles big data Schema-free NoSQL types Flexibility, speed NoSQL tradeoffs Jinadu et al. (2021)[45] DSP architecture RAID, MTCP Faster response Mobilefriendly M-TCP, WLAN Response time Coordination load Hongwei & Ligetu (2021)[46] Cloud storage Object store Low-cost storage Adaptable Cloud, object store Cost, latency Integration limits Chang & Cui (2021)[47] Economic data Geo+Hilbert Efficient storage Cross-modal Cassandra, Hilbert Storage use Model tuning Thamar (2023)[48] Data pipelines Modern tools Real-time boost Streamlined Kafka, Spark Real-time ops Tool mix Zhang et al. (2024)[49] Anomaly detect LSTM + MultiLog F1 12% Multi-node accurate 900M log entries F1 score Single-node limits Olusegun et al. (2024)[50] Secure analytics SMPC, FL Safe collaboration Cloud-ready Multi-cloud Security Overhead Munawar et al. (2024)[51] Smart apps 7Vs, IoT Helps cities/disasters Flexible IoT, social media Process time Real-time limits Ibrahim (2024)[52] Data sync Dsync logs No timestamp needed Scalable Dsync logs Sync speed Metadata sync Gadde (2024)[53] AI in DB ML + anomaly Violations 80% AI boosts integrity Cloud sim. Throughput ML overhead Aryan et al. (2024)[54] DADBS Rust + PoW 3000 TPS Resilient Rust, 500 nodes TPS, consistency Latency, PoW Zhu et al. (2025)[55] Redis optimize RAPO, LBI Latency 30% Balanced reads Redis cluster Latency drop Write scaling Sato (2025)[56] DB roles Cloud/micro Modern DB shift Architect role Architectural review Design agility Skill gaps Kirino (2025)[57] DB design Crossdomain Governance aware Balanced Design concepts Observability Tool balance Evelyn (2025)[58] Query opt. ML + metadata Query time 40% Adaptive PostgreSQL, Hive Execution time Source diversity Ailamaki (2025)[59] Query exec Spark, Dask Latency improved Self-tuning Spark, Dask Latency, fault tol. Cost model gaps Adeleke (2025)[60] Approx. algo Sampling, sketching Fast queries Efficient Spark, HDFS Response, error Coordination Abiteboul et al. (2025)[61] Query optimization Cost, ML, quantum Performance Hybrid models Simulated env. Query time Quantum/ML 4. Table 1. Comparison among the reviewed works Synthesizing 22 empirical and conceptual studies spanning 2020 to 2025. “A Survey on Distributed Database Systems in the Era of Big Data” 7715 ETJ Volume 10 Issue 11 November 2025, 1 Kazheen Ismael Hasan 5. DISCUSSIONS & COMPARISON Table 1 provides a detailed summary of the previous publications. It presents the key indicators used for assessment and emphasizes the significant findings from these studies, highlighting the strengths and innovative concepts that emerged from the research. Synthesis of reviewed paper The examined literature illustrates a dynamic and evolving domain characterized by recurring themes of performance enhancement, scalability, and system adaptability. A thorough analysis indicates that although basic distributed architectures are firmly established, research is increasingly focused on enhancing efficiency, integrity, and intelligent automation. Performance optimization continues to be a primary emphasis.[40] illustrated that Riak KV performs exceptionally well in read-dominant workloads but falters in update-intensive scenarios, particularly with substantial datasets, highlighting the constraints of linear scalability. [45] and [55] proposed DSP and RAPO architectures that markedly enhance throughput and latency, highlighting performance optimization via virtualization and node load balancing. Incorporating AI and machine learning in query optimization signifies a pivotal transition towards adaptive and autonomous systems. [58] utilized machine learningbased cost models, resulting in a 40% enhancement in query time. [53]developed an AI-driven anomaly detection model that decreased transaction violations by 80%, demonstrating the efficacy of predictive analytics in improving transactional integrity. Scalability and fault tolerance are paramount. [61]identified the shortcomings of traditional MapReduce. They proposed a Distributed Hash Table (DHT)-based system utilizing Poincaré disk geometry, thereby improving decentralized coordination and mitigating single points of failure. [54]further developed a Rust-based decentralized database that can manage 3,000 transactions per second (TPS) and exhibits near-linear scalability up to 500 nodes. Security and data governance, while less commonly examined, are becoming vital issues.[50]examined Secure Multi-Party Computation (SMPC) in big data analytics, demonstrating that privacy-preserving protocols are viable, though accompanied by communication overhead.[52]addressed synchronization in heterogeneous environments by introducing "Dsync logs," which obviate the need for timestamps in real-time, bi-directional data updates. The literature reveals an imbalance: Although system performance is extensively studied, topics like data ethics, secure collaboration, and governance receive significantly less empirical focus despite their growing importance in distributed, multi-cloud environments. This view indicates a necessity for more integrative strategies that harmonize technical efficiency with ethical and operational resilience [56],[57] In this manner, the trajectory of distributed database research demonstrates a comprehensive understanding of performance mechanics yet exposes significant deficiencies in aspects such as cross-platform interoperability, secure collaboration, and long-term governance. Future initiatives must progress beyond optimization to adopt resilient, ethical, and intelligent architectures that satisfy the requirements of modern data ecosystems. 6. EXTRACTED STATISTICS Figure (1) depicts a multifaceted research landscape in distributed databases and big data systems, emphasizing diverse focal points without a predominant theme. The persistent interest in query optimization, AI integration, and NoSQL significantly indicates a collective emphasis on improving performance and scalability. As demonstrated, the focus on machine learning-based optimization signifies a transition towards intelligent, self-adaptive data systems. Simultaneously, underrepresented domains such as secure analytics and data synchronization present significant privacy, consistency, and system interoperability issues, highlighting these subjects' complexity and evolving nature. The insufficient focus on intelligent applications and economic data storage indicates a disparity between research and sector-specific requirements. The distribution indicates a domain that harmonizes innovation with execution. Future research must integrate theoretical insights with practical limitations, ensuring that distributed systems are efficient, ethical, secure, and contextually aware. Figure 1 Statistical representation of Big Data papers (2020 – 2025) based on the focus area Figure (2) represents the thematic allocation of principal findings from the analyzed research on distributed databases and big data systems. Most research focuses on enhancing performance, indicating the academic community's emphasis on speed, throughput, and system responsiveness in progressively intricate data environments. The significant emphasis on query optimization and latency reduction reflects continuous endeavors to enhance data access efficiency and timeliness. Simultaneously, real-time functionality and storage optimization demonstrate the demand for scalability and agility in dynamic environments such as IoT and cloud platforms. Infrequent yet equally significant are themes such as privacy and security, data synchronization, and governance, which evoke ethical and technical issues frequently neglected in performance-focused discussions. The chart highlights the necessity for a comprehensive research methodology that improves system performance while tackling integrity, coordination, and ethical data utilization in distributed architectures. “A Survey on Distributed Database Systems in the Era of Big Data” 7716 ETJ Volume 10 Issue 11 November 2025, 1 Kazheen Ismael Hasan Figure 2 Statistical representation of Big Data papers (2020 – 2025) based on the key findings Figure (3) illustrates a thematic analysis of the benefits identified in research on distributed databases and big data systems. Scalability is highlighted as the paramount advantage, signifying the urgent need to manage increasing data volumes and effectively dispersed workloads. The subsequent themes pertain to efficiency and performance optimization, illustrating continuous endeavors to refine system operations, minimize latency, and improving responsiveness. Flexibility underscores the significance of adaptable systems that accommodate various data types and changing requirements. Infrequent but essential benefits, including security, accuracy, and governance, indicate an increasing recognition of ethical and operational intricacies. The distribution indicates a research landscape predominantly focused on technical performance objectives while becoming progressively mindful of adaptability and trust. In practical settings, future development must prioritize the equilibrium between system optimization and overarching issues such as user autonomy, data ethics, and cross-platform resilience. “A Survey on Distributed Database Systems in the Era of Big Data” 7717 ETJ Volume 10 Issue 11 November 2025, 1 Kazheen Ismael Hasan Figure 3 Statistical representation of Big Data papers (2020 – 2025) based on the Advantages Figure (4) displays the varied performance metrics highlighted in the literature examined on distributed databases and big data systems. Latency and throughput are the most addressed metrics, highlighting the field's persistent focus on speed and responsiveness. Query time is significant, indicating the necessity for efficient data retrieval in real-time and extensive environments. Efficiency, real-time operations, and system throughput exemplify the diverse objectives of optimizing computational resources and user experience. Thus, underrepresented yet vital aspects such as synchronization, security, and monitoring highlight systemic issues that, while frequently subordinate to performance, are crucial for enduring stability and trust. This distribution indicates a performance-oriented paradigm in contemporary research; however, it also prompts contemplation: as systems become increasingly interconnected and data-intensive, there is an escalating necessity to reconcile speed with reliability, transparency, and adaptability in distributed data architectures. Figure 4 Statistical representation of Big Data papers (2020 – 2025) based on the Performance 7. RECOMMENDATIONS & FUTURE AREA A fundamental recommendation is to incorporate ethics by design into their architecture to promote the responsible development of distributed database systems to promote the responsible development of distributed database systems. Contemporary development practices frequently emphasize “A Survey on Distributed Database Systems in the Era of Big Data” 7718 ETJ Volume 10 Issue 11 November 2025, 1 Kazheen Ismael Hasan technical performance while neglecting inherent ethical considerations, including fairness, transparency, and user autonomy. Future systems, especially those utilizing AI for autonomous decision-making, must incorporate ethical considerations from the initial phases of system modelling and design [57][53]. Moreover, performance should not be sought in isolation. Distributed databases must adequately facilitate governance, auditability, and interoperability, especially within multi-tenant or federated frameworks. This view involves implementing synchronization mechanisms, data lineage tracking, and compliance-ready features as standard components [52],[50]. Furthermore, an increasing dependence on databases in essential services necessitates that resilience and decentralization be considered fundamental design principles. Decentralized Autonomous Databases (DADBS) and Distributed Hash Table (DHT) architectures present effective frameworks for fault tolerance and operational continuity in adverse conditions [61]&[55],[54] Finally, creating multi-faceted evaluation frameworks that extend beyond conventional latency or throughput metrics is essential. These must encompass energy consumption metrics, explainability, user autonomy, and regulatory compliance—ensuring that distributed systems are efficient, sustainable, and socially responsible. Ultimately, distributed databases' future must be engineered and ethically designed. As these systems increasingly support essential societal infrastructures, from healthcare and finance to public administration, the stakes have transcended mere technical considerations. Researchers and practitioners must adopt multi-dimensional design methodologies integrating performance, ethical foresight, regulatory compliance, and civic responsibility. This outlook involves transitioning from discrete technological enhancement to system thinking, where databases are more efficient and intelligent but equitable, secure, and transparent. This transition is not merely advantageous, it is essential. Future research should primarily focus on enhancing the scalability of privacypreserving computation. With the increasing rigor of data sovereignty and cross-border regulations, it is imperative to adapt mechanisms such as Secure Multi-Party Computation (SMPC) and Federated Learning for large-scale, low-latency environments, particularly in multi-cloud deployments [50]. A promising avenue is the advancement of explainable AI for database optimization. As AI-driven systems progressively execute autonomous decisions regarding query planning, anomaly detection, and integrity verification, researchers must guarantee that these decisions are interpretable and auditable for technical users and governance entities [58],[53]. Hence, cross-domain applications of distributed databases are still inadequately investigated. The practical application in healthcare, public administration, and smart city infrastructures would evaluate the proposed models' resilience and underscore the conflicts between technical scalability and ethical accountability. Research must examine these intersections to guarantee that distributed systems are adaptable to socially sensitive contexts. The environmental impact of distributed architectures must be prioritized. The energy intensity of replication protocols, consensus mechanisms, and continuous synchronization operations must be examined in the context of overarching objectives of green computing and sustainable data infrastructure. Overrepresentation of performance-centric studies is a major limitation. These contributions improve our understanding of scalability and efficiency but often sacrifice socio-technical dimensions. Thus, algorithmic bias, data ethics, and end-user empowerment are understudied. Many proposed frameworks, such as RAPO, DSP, and DADBS, are validated only in simulation or controlled environments, raising questions about their real-world viability[55], [45]. Geographic and infrastructural bias in much of the reviewed literature is another limitation. Few studies examine how distributed databases work in low-resource or Global South contexts, where bandwidth, electricity, and regulations vary. Existing solutions are less global and inclusive due to this neglect. User-centered evaluation is scarce in the field. What humans do with distributed systems, how they interpret outputs, how much they trust automated processes, and how system opacity affects decision-making have received little attention. Addressing these limitations is crucial to creating technically robust and socially legitimate distributed databases. 8. CONCLUSIONS In the big data age, distributed database systems have rapidly evolved with a primary focus on improving performance measures, such as query response, latency, and throughput. Even if these advancements are essential for mission-critical and high-volume applications, placing too much focus on computing speed runs the danger of encouraging a technocentric viewpoint that ignores crucial elements like system transparency, architectural complexity, and environmental impact. Incorporating economic, environmental, and human-centered measures into performance reviews would be a more equitable strategy. A major change has been made to database systems with the incorporation of artificial intelligence (AI) and machine learning (ML), which allows for adaptive query optimization and real-time anomaly identification. However, this advancement raises issues with possible bias, ethical governance, and algorithmic transparency. Addressing concerns of accountability and interpretability becomes essential as these systems transform from passive instruments to proactive decision-makers. Despite progress, there is still a significant gap in the areas of governance-focused design, data synchronization across heterogeneous systems, and privacy-preserving computing. Especially in multi-cloud and global data settings, these elements are essential for guaranteeing trust, interoperability,