Performance comparison of microservices written using reactive and imperative approaches

Abstract

The purpose of this paper is to compare the performance of microservices based on reactive and imperative approaches. For this purpose, two microservice applications written in Java using the Spring programming framework were developed. The Spring Web and Spring Webflux modules were used for the conventional and reactive versions, respectively. During the tests, functionalities related to operations of retrieving and inserting records into the database, data processing, and file transfer were invoked. The Gatling tool was used to conduct the tests. The tests showed that reactive microservices can be more efficient in particular when there are delays in communication with services or the database. Otherwise, it depends on the complexity of the operations being performed. Microservices based on the reactive paradigm also use less RAM compared to conventional counterparts.