Rcaeval

RCAEval is a comprehensive benchmark, consisting of three RCA datasets and a comprehensive evaluation environment with fifteen baselines (https://arxiv.org/abs/2412.17015). First, our datasets include 735 failure cases collected from three microservice systems, covering 11 fault types observed in real-world failures. We collected multi-source telemetry data (i.e., metrics, logs, and traces), supporting a variety of RCA approaches (e.g., metric-based, trace-based, multi-source RCA). Second, we release our evaluation framework as an open-source library, which includes reproducible RCA baselines for benchmarking. The prior version of this framework was used to evaluate metric-based RCA (https://dl.acm.org/doi/10.1145/3691620.3695065). In this work, we have upgraded it to support trace-based and multi-source RCA.

RCAEval is a comprehensive benchmark, consisting of three RCA datasets and a comprehensive evaluation environment with fifteen baselines (https://arxiv.org/abs/2412.17015). First, our datasets include 735 failure cases collected from three microservice systems, covering 11 fault types observed in real-world failures. We collected multi-source telemetry data (i.e., metrics, logs, and traces), supporting a variety of RCA approaches (e.g., metric-based, trace-based, multi-source RCA). Second, we release our evaluation framework as an open-source library, which includes reproducible RCA baselines for benchmarking. The prior version of this framework was used to evaluate metric-based RCA (https://dl.acm.org/doi/10.1145/3691620.3695065). In this work, we have upgraded it to support trace-based and multi-source RCA.

RCAEval is a comprehensive benchmark, consisting of three RCA datasets and a comprehensive evaluation environment with fifteen baselines. First, our datasets include 735 failure cases collected from three microservice systems, covering 11 fault types observed in real-world failures. We collected multi-source telemetry data (i.e., metrics, logs, and traces), supporting a variety of RCA approaches (e.g., metric-based, trace-based, multi-source RCA). Second, we release our evaluation framework as an open-source library, which includes reproducible RCA baselines for benchmarking. The prior version of this framework was used to evaluate metric-based RCA (https://dl.acm.org/doi/10.1145/3691620.3695065). In this work, we have upgraded it to support trace-based and multi-source RCA.

RCAEval is a comprehensive benchmark, consisting of three RCA datasets and a comprehensive evaluation environment with fifteen baselines. First, our datasets include 735 failure cases collected from three microservice systems, covering 11 fault types observed in real-world failures. We collected multi-source telemetry data (i.e., metrics, logs, and traces), supporting a variety of RCA approaches (e.g., metric-based, trace-based, multi-source RCA). Second, we release our evaluation framework as an open-source library, which includes reproducible RCA baselines for benchmarking. The prior version of this framework was used to evaluate metric-based RCA (https://dl.acm.org/doi/10.1145/3691620.3695065). In this work, we have upgraded it to support trace-based and multi-source RCA.

This repository contains the Software Artifact for reproducing the main experimental results in our paper accepted to ASE 2024: "Root Cause Analysis for Microservice System based on Causal Inference: How Far Are We?".

The artifact is also available in the Github repository: https://github.com/phamquiluan/RCAEval.

This repository contains the Software Artifact for reproducing the main experimental results in our paper accepted to ASE 2024: "Root Cause Analysis for Microservice System based on Causal Inference: How Far Are We?".

The artifact is also available in the Github repository: https://github.com/phamquiluan/RCAEval.