CARE: Context Aware Root Cause Identification Using Distributed Traces and Profiling Metrics

Abstract

Root cause localization in microservices is challenging due to intricate service dependencies and the high volume and heterogeneity of collected monitoring data, which add complexity to the analysis. Conventional methods often overlook nuanced propagation patterns and contextual interactions among services, and they are limited in leveraging multi-source observability data for comprehensive root cause identification. This study introduces CARE, a context-aware, spectrum-analysis-based approach that integrates multi-source observability data and employs network analysis to prioritize the contextual significance of components in propagating anomalies across individual services, service communities, and requests. CARE's weighted spectrum analysis leverages these prioritized contexts to pinpoint underlying performance issues. Evaluations on 224 cases from the TrainTicket benchmark and a real-world Internet service provider's production system demonstrate CARE's substantial accuracy gains, with top-1 accuracy of 72%-89% and top-5 accuracy of 84%-99% for single root causes, outperforming baselines by 8%-41%. CARE also shows significant improvements in dual root cause identification, exceeding baseline performance by 18%-37%, all while maintaining efficient resource usage, establishing CARE as a robust and resource-effective solution for root cause localization in complex microservice environments.

Source: Exact abstract from the public author-accepted manuscript in Zenodo record 18021268, verified on 2026-08-09.

Problem and motivation

Root-cause localization in microservices is difficult because service dependencies, request paths, execution context, and heterogeneous monitoring data interact; conventional methods can miss propagation patterns and contextual relationships.

Method and contribution

CARE builds service-call graphs from distributed traces, detects anomalous profiling metrics, clusters affected traces, ranks the contextual importance of services and service communities with graph analysis, and applies weighted spectrum-based fault localization. The evaluation compares regular spectrum analysis, MicroRank, TraceRCA, and HeMiRCA using Ochiai, M2, DStar2, and Russell-Rao risk formulas.

Findings and evidence

On TrainTicket, which contains 41 microservices and 242,259 traces across 200 fault scenarios, CARE reports top-1 accuracy of 79%-89% and top-5 accuracy of 94%-99%. On the China Mobile Zhejiang production dataset, the reported top-1 accuracy is 68%-72% and top-5 accuracy is 84%. For 11 double-root-cause scenarios, both causes are placed in the top two in 55% of cases and in the top five in 82% of cases.

Limitations and future directions

Limitations: The evaluation depends on the selected TrainTicket and China Mobile Zhejiang datasets and their fault-injection or anomaly-selection procedures. The paper also reports a mismatch for HeMiRCA on the production data because of mixed normal/anomalous traces, sparse anomalies, and missing caller-side metrics for some services.

Future work: Extend validation to broader production systems and more complex multi-root-cause scenarios, with richer context modeling and online diagnosis under trace sampling.

Sources and identifiers

• Published version published
• Accepted-author manuscript · PDF public_full_text

When to cite this paper

Cite this paper when combining distributed traces, profiling metrics, and context-aware spectrum analysis for microservice root-cause localization.

• Graph-based ranking of service and community context from distributed traces.
• Weighted spectrum-based localization using profiling anomalies and multiple risk formulas.
• TrainTicket and China Mobile Zhejiang evaluations, including single- and double-root-cause scenarios.

Citation

@article{ezzatiJivan2026carecontext,
  author = {Mahsa Panahandeh and Naser Ezzati-Jivan and Abdelwahab Hamou-Lhadj and James Miller},
  title = {CARE: Context Aware Root Cause Identification Using Distributed Traces and Profiling Metrics},
  year = {2026},
  journal = {IEEE Transactions on Software Engineering},
  volume = {52},
  number = {2},
  pages = {691-715},
  publisher = {Institute of Electrical and Electronics Engineers (IEEE)},
  issn = {0098-5589, 1939-3520, 2326-3881},
  doi = {10.1109/tse.2025.3645143},
  url = {https://doi.org/10.1109/TSE.2025.3645143}
}

Panahandeh, M., Ezzati-Jivan, N., Hamou-Lhadj, A., & Miller, J. (2026). CARE: Context Aware Root Cause Identification Using Distributed Traces and Profiling Metrics. IEEE Transactions on Software Engineering, 52(2), 691-715. https://doi.org/10.1109/tse.2025.3645143

M. Panahandeh, N. Ezzati-Jivan, A. Hamou-Lhadj, and J. Miller, "CARE: Context Aware Root Cause Identification Using Distributed Traces and Profiling Metrics," IEEE Transactions on Software Engineering, vol. 52, no. 2, pp. 691-715, 2026, doi: 10.1109/tse.2025.3645143