When designing microservice systems, the choice of a technological platform requires a reliable and comparable assessment of its performance under real workload conditions. The subject of this study is the experimental analysis of microservice interaction performance implemented using different web frameworks. The object of the research is server-side technologies used in the development of microservice applications. The relevance of the study is determined by the need for objective and comparable performance evaluation when selecting a technological platform for microservice architectures under realistic load conditions. Special attention is given to ensuring the comparability of results by keeping the business logic, service interaction scenarios, and infrastructure environment parameters unchanged. The study includes a comparison of three implementations of a microservice application developed using ASP.NET Core, NestJS, and Gin. The evaluation was performed using the following metrics: response time, p95 latency, throughput (RPS), CPU utilization, and memory consumption. The research was conducted using an applied experimental approach based on a containerized Docker infrastructure. Load testing was performed using the k6 tool, while resource consumption monitoring was carried out using Prometheus and cAdvisor under fixed experimental conditions. The main outcomes of the study are the results of a comparative performance analysis of microservice applications implemented with different web frameworks. The Gin-based implementation demonstrated the lowest response time and p95 latency, minimal computational resource consumption, and the highest throughput. ASP.NET Core provided stable performance with moderate CPU utilization but showed increased memory consumption. NestJS demonstrated the highest latency values and CPU load. The scientific novelty of the study lies in the development and formalization of a reproducible methodology for comparative evaluation of microservice interaction performance, taking into account both time-related characteristics and resource consumption metrics. The practical significance of the results lies in the possibility of using the proposed methodology to support informed technology stack selection and architectural decision-making in the development of distributed systems.
Monichev et al. (2026) studied this question.