Fault Localization Using Interventional Causal Learning for Cloud-Native Applications

Abstract

In this work, we share our experience of using recently proposed fault localization techniques based on interventional causal learning applied in the context of cloud-native applications. We identify several assumptions that prevent successful deployment of interventional causal learning fault localization in the real-world. These assumptions directly contradict the established knowledge-base of the current causality-driven work in this domain. Based on those insights, we make the following contributions: (i) development of minimal benchmark application, CausalBench, that surfaces these challenges, and (ii) adapting the interventional causal learning technique using system insights to address these challenges. Our fault localization work outperforms the recent state-of-the-art algorithms.