Group Project Proposal$

Bowen Liang, YISEN HONG, Chenghao Li

Keywords: microservice, root cause analysis

Abstract:

Root cause analysis (RCA) in microservice systems is challenging because a single fault can trigger many correlated downstream abnormalities, making it difficult to distinguish causes from symptoms. Existing methods improve localization accuracy but often lack explicit support for verifying why a predicted root cause should be trusted. In practice, operators require predictions that can plausibly explain the rest of the incident.

This work proposes a verifiable RCA framework that combines temporal evidence learning with root-centered partial causal graphs. Each candidate is evaluated by both its local evidence and its ability to explain downstream effects through structured propagation. The resulting graph serves as both a scoring mechanism and a faithful explanation, improving the robustness and practical usefulness of RCA.

Submission Number: 7