Serverless & Disaggregated Database Architectures: A Systematic review
TL;DR
Key findings reveal that serverless and disaggregated databases enable elastic scaling and significant cost reductions through pay-as-you-go models and resource pooling, while performance optimization benefits from AI-driven scheduling and hardware co-design; however, challenges persist in cold-start latency, orchestration complexity, and consistency management.
Abstract
Review synthesizes research on "Serverless and disaggregated database architectures: performance optimization, cost efficiency, scalability, historical evolution, current trends, practical applications in industries, comparison with traditional database systems" to address the knowledge gap regarding how emerging paradigms reshape database management in cloud-native environments. The review aimed to taxonomies architectural designs, evaluate performance and cost-efficiency strategies, benchmark scalability, identify industry applications, and compare historical evolution with current trends. A systematic analysis of empirical studies, prototypes, and theoretical works published up to mid-2024 was conducted, focusing on cloud-native deployments leveraging technologies such as RDMA, persistent memory, and function orchestration. Key findings reveal that these architectures enable elastic scaling and significant cost reductions through pay-as-you-go models and resource pooling, while performance optimization benefits from AI-driven scheduling and hardware co-design; however, challenges persist in cold-start latency, orchestration complexity, and consistency management. Industry adoption spans finance, retail, and IoT, demonstrating operational gains but constrained by migration complexity and tooling maturity. The evolution from monolithic to serverless and disaggregated systems is marked by innovations in decoupled resource management and multi-cloud strategies. These findings collectively underscore the transformative potential and practical limitations of serverless and disaggregated databases, informing future research and guiding effective industrial adoption.