Experimental Analysis of Performance and Resource Stability in Cloud-Native Microservices

Abstract

The widespread adoption of microservices architecture has enabled modern technology organizations to achieve greater system flexibility and scalability. However, this shift introduces significant challenges in managing performance and resource stability due to the system's inherently distributed nature. In this study, we evaluate the dynamic behavior of a microservices system, using the Google Online Boutique project as a representative case. Our methodology involved a series of systematic tests, including baseline, stress, and long-duration soak tests, executed within a Docker container environment. We monitored performance using tools such as JMeter and Docker stats. Our experiments identified a maximum system throughput of 115.5 requests per second. Crucially, we observed that exceeding this threshold triggered a nonlinear degradation in response time, which escalated to 2016ms. The analysis pinpointed a computational bottleneck within the Currency Service. Furthermore, the soak test revealed a 12% increase in memory consumption over time, indicating a potential resource leak. These findings underscore the critical importance of Horizontal Scaling strategies and proactive resource management in cloud-native environments.