Migration from Monolith to Microservices: Benchmarking a Case Study.

Abstract

Migrating from a monolith architecture to a mi-croservice architecture in order to modernize a system seems to have become popular in the recent years. However, the benefits of this migration does not appear to have been sufficiently verified. The goal of this article is to present a methodology able to verify whether or not a migration to microservices is beneficial. The methodology is derived starting from a systematic literature review useful to survey and analyze existing approaches in the field and to identify a set of metrics (i.e., latency, throughput, scalability, CPU usage, memory usage, and network usage) needed to benchmark the source and the target applications. We evaluate the solution proposed conducting benchmarking experiments on a reference system we have developed in two versions, monolith and microservices. We conduct small-scale experiments using consumer-grade hardware, and large-scale experiments using cloud services.