Performance and Resilience Impact of Microservice Granularity: An Empirical Evaluation Using Service Weaver and Amazon EKS

Determining the optimal granularity level for microservices applications is a critical challenge in modern software architecture. This study leverages the Service Weaver framework to investigate the performance and resilience implications of different service granularity configurations in a public cloud environment. We deployed multiple configurations of the Online Boutique microservice demo application on Amazon Elastic Kubernetes Service (EKS) and conducted a series of experiments to evaluate their behavior under varying workloads and failure conditions. Our results indicate that distributing services across multiple EKS nodes can significantly enhance scalability, particularly under high workloads, but at the cost of increased communication overhead. We also found that while cloud-native resilience mechanisms, such as automatic re-starts and retries, effectively mitigate frequent random failures, they tend to impose a notable performance overhead, especially in configurations with tightly coupled services. Our findings highlight the importance of carefully balancing service granularity with both performance and resilience considerations when designing robust cloud-based microservice applications.