Performance Impact on Databases Using Serverless Architectures: An Empirical Study

Abstract

Serverless computing is rapidly gaining traction as a viable alternative for deploying and managing various applications. Serverless databases, in particular, promise scalability, cost-effectiveness, and reduced operational overhead. However, the performance characteristics of serverless databases, especially in comparison to traditional database deployments, are not yet fully understood. This paper presents an empirical study investigating the performance impact of utilizing serverless database architectures. We conduct experiments using a benchmark workload against both a serverless database (AWS Aurora Serverless v2) and a provisioned database (AWS Aurora PostgreSQL). Our results reveal that while serverless databases offer significant benefits in terms of scalability and cost, they can introduce latency overhead due to cold starts and autoscaling mechanisms. We analyze these performance nuances, providing insights into the trade-offs between serverless and provisioned database deployments for various application scenarios