Uma análise comparativa de desempenho entre diferentes tecnologias de execução de aplicações web do lado do servidor

Abstract

This work compares the performance of native applications against applications to be executed by execution systems. The applications were built with two frameworks meant for server-side application development with Java: Spring Boot and Quarkus. It was built an application for each framework. Both applications were compiled to native code with GraalVM and to jar packages to be executed by JVM. The adopted criterion to check the performance was the average response time from server to client for each created scenario, it was also evaluated the initialization time of each application. Both applications work like APIs that access a database. It was also created a method to sort a vector of numbers passed by the HTTP request body. It was used two machines to work as server and client alternately, the server executed the application while the client accessed the server and registered the results statistics. The tables of registered results are shown in this work. It was noted that in most of the tested cases the Quarkus native application had a reduction in average response time. To Spring Boot there wasn’t any confirmation of performance improvement to the native application. The applications were also encapsulated in Docker to observe how they perform in this circumstance. There wasn’t a worsening observed in the encapsulated version performance, although this result isn’t conclusive.