Avaliação de desempenho de infraestrutura serverless

Abstract

Serverless computing has gained widespread adoption due to its simplified management and lightweight design, especially when combined with containerized systems. This model allows developers to focus exclusively on application logic, eliminating the need to manage the underlying infrastructure. The serverless model offers advantages such as billing based on execution time, in millisecond units, reducing operational costs and attracting companies interested in greater efficiency. This study investigates the resource usage in serverless environments using Knative, evaluating different combinations of container platforms and operating systems (Ubuntu/Docker, Ubuntu/Podman, Debian/Docker and Debian/Podman), under workloads simulating 100 users (low load), 500 users (medium load) and 1000 users (high load). In addition, it also investigates the effects of software aging under workload stress. The objective is to propose infrastructure configurations that maximize efficiency in resource use while maintaining good application performance levels. The results demonstrate that the Ubuntu/Docker combination presents better efficiency in resource allocation in performance tests, while in software aging tests the absence of progressive degradation in performance metrics reinforces the potential of serverless architectures to resist the effects of software aging through automatic scaling. The conclusions provide insights for optimizing resource use in open source serverless platforms, contributing to better decisions in production environments.