DeLag: Using Multi-Objective Optimization to Enhance the Detection of Latency Degradation Patterns in Service-Based Systems

Performance debugging in production is a fundamental activity modern service-based systems. The diagnosis of performance issues often time-consuming, since it requires thorough inspection large volumes traces and indices. In this paper we present DeLag, novel automated search-based approach for diagnosing DeLag identifies subsets requests that show, the combination their Remote Procedure Call execution times, symptoms potentially relevant issues. We call such Latency Degradation Patterns. simultaneously searches multiple latency degradation patterns while optimizing precision, recall dissimilarity. Experimentation on 700 datasets generated from two microservice-based systems shows our provides better more stable effectiveness than three state-of-the-art approaches general purpose machine learning clustering algorithms. effective all baseline techniques at least one case study (with p $\leq$ 0.05 non-negligible effect size). Moreover, outperforms terms efficiency second third most largest used evaluation (up to 22%).