Acceso usuarios
The demand for digital transformation has led companies to adopt solutions that promote automation, integration, and scalability of services using APIs and cloud platforms. This study evaluates two scenarios that favor the use of blocking versus non-blocking REST APIs, considering performance metrics and resource consumption (response time, CPU usage, and RAM consumption). Applications were developed for the experiment using Spring Boot (blocking) and Spring WebFlux (non-blocking): product listing and video streaming. Load testing was conducted with JMeter in incremental scenarios ranging from 10 to 1000 simultaneous requests. Statistical analysis employed non-parametric tests (Mann-Whitney U and Brunner-Munzel). Spring WebFlux demonstrated greater efficiency in high-demand scenarios (lower response times [0.84ms vs. 3414ms in streaming] but with higher RAM consumption [2.24GB vs. 0.88GB] and more frequent errors [OutOfMemoryError] under heavy loads [750 errors vs. 120 in 1000 requests], optimizing thread usage and improving scalability). However, its structural complexity was higher, requiring greater attention to asynchronicity. Spring Boot stood out for greater stability (average CPU usage of 24.06% vs. 23.36% in WebFlux), lower memory consumption (0.88GB vs. 2.24GB in streaming), and zero errors (0 vs. 750 errors in critical scenarios), maintaining consistent performance even under high concurrency, making it more suitable for low-concurrency applications. The analysis revealed that both approaches have advantages and limitations, and the choice should consider factors such as request volume and performance requirements.
