Comparação quantitativa entre requisições de diferentes bibliotecas: um estudo comparativo de desempenho de bibliotecas HTTP em Python, Node.js, Go e Elixir

Abstract

The modern web development landscape demands increasingly performant and scalable applications to handle a growing volume of concurrent users. The choice of backend technology, especially the libraries responsible for network communication, becomes a critical factor that directly impacts a system's responsiveness. This work conducts a quantitative comparison of the throughput performance of eight HTTP request libraries across four popular programming languages: Python (requests, httpx), Node.js (axios, undici), Go (net/http, fasthttp), and Elixir (HTTPoison, Finch). Through a controlled experimental methodology using Docker containers and a stress-testing scenario focused on I/O operations, the maximum capacity of each library to process requests per minute was measured. The results demonstrated a significant superiority of the Elixir and Go libraries, which leverage their native concurrency models to achieve substantially higher throughput rates. In contrast, the Python libraries showed the greatest scalability limitations. The study concludes that the adoption of asynchronous programming paradigms and the choice of ecosystems with efficient concurrency support are fundamental for developing high-performance network applications.