PRiskLoc: An Enhanced Multi-dimensional Root Cause Localization Algorithm Aided by a Fine-Grained Evaluation System

Root cause localization within multi-dimensions is a challenging task due to its large search space within a limited time. There are a series of algorithms to handle this task, but to our knowledge, there is no evaluation system to help users analyse or optimize them according to their specific data and needs. In this paper, there are two main contributions: first, we provide a multi-dimensional evaluation system to evaluate the performance of algorithms in full aspects, which can help us comprehensively and finely analyse, compare, and choose the applicable scenario of algorithms or optimize targeted algorithms; second, we analyse and find the weakness of the SoTA algorithm RiskLoc based on our contributed evaluation system, aiming at its weakness. To tackle the issue of RiskLoc found by our evaluation system, we present PRiskLoc, an efficient and effective multi-dimensional root cause localization algorithm. We demonstrate that PRiskLoc consistently outperforms state-of-the-art baselines, especially in more challenging root cause scenarios, with the F1 improved from 0.635049 to 0.724687.