Breaking monolithic applications into microservices is well studied, but the redesign efforts often stop after the initial decomposition. However, microservice architectures evolve, and changing requirements demand ongoing effort to optimize and reduce the interservice communication to provide the best possible performance. Nevertheless, redesigning an existing system is challenging due to established domain or functional boundaries that limit flexibility. To address this issue, we present a novel approach for refining existing microservice architectures to reduce communication overhead while preserving original domain and data access constraints. Our method applies a community detection algorithm, guided by forces derived from domain boundaries, data consistency, and functional separation, to identify optimal service clusters. By running our algorithm with varying input scenarios, we generate a Pareto front of system redesign alternatives, evaluated on architectural metrics. In a case study using a large microservice reference system, our approach reduced interservice calls by 20% while keeping all constraints and up to 50% when partially easing some service boundaries. The approach is easily configurable and adaptable, offering a practical tool for evolving microservice architectures.

Top