A Quantitative Analysis of Serverless Functions Against Microservice Clusters

Abstract

The evolution of cloud computing architectures has introduced two dominant paradigms for application deployment: serverless functions and microservice clusters. Organizations face critical decisions when selecting between these approaches, yet comparative quantitative analysis remains limited. This research provides a comprehensive empirical comparison of serverless functions and microservice clusters across multiple performance dimensions including latency, scalability, cost-efficiency, and resource utilization. Through controlled experiments deploying identical workloads on AWS Lambda (serverless) and Kubernetes-based microservices, this study quantifies the trade-offs between these architectural patterns. Testing involved three application scenarios—a REST API service, an event-driven data processing pipeline, and a web application backend—each subjected to varying load conditions. Results indicate that serverless functions demonstrate superior automatic scaling capabilities and cost-efficiency for sporadic workloads, reducing costs by 40-65% compared to microservices for low-utilization scenarios. However, microservice clusters outperform serverless in sustained high-load conditions, exhibiting 35-50% lower latency and greater predictability. The cold start penalty for serverless functions averaged 850-1200ms, significantly impacting user-facing applications. This research provides quantitative evidence to guide architectural decisions based on workload characteristics, contributing empirical data to the ongoing debate between these competing paradigms.