Extracting Microservices’ Candidates from Monolithic Applications: Interface Analysis and Evaluation Metrics Approach

Abstract

There is a migration trend toward microservices architecture coming from the monolithic applications. This research proposes a decomposition method that extracts microservices’ candidates through analyzing the application programming interface in order to extract the operations and the parameters. Then the operation names are converted into word representations using word embedding models. Next, semantically similar operations are clustered together to provide a microservice’ candidate. Additional step is to evaluate the proposed candidate using cohesion and complexity metrics. The proposed algorithm improved the decomposition approach for big applications but did not affect the decomposition of smaller applications.