The DOI matches its Crossref record; 203 of the 204 references it cites were found in a registry or an index; 42 of the 210 sources its claims cite were checked for support.
Anomaly detection and root-cause identification in microservices: a survey
Open accessAbstract
Microservices Architecture (MSA) has emerged as a dominant paradigm for building scalable and flexible distributed systems. By decomposing applications into loosely coupled services, MSA improves modularity and deployment agility, but also introduces significant challenges for system monitoring, anomaly detection, and root-cause identification. These challenges stem from the high degree of dynamic behaviour, service heterogeneity, and inter-service dependencies. This survey provides a comprehensive and systematic review of state-of-the-art approaches for anomaly detection and root-cause analysis in microservice-based systems, covering the period from 2012 to 2025. It examines 117 selected studies, categorized according to data collection methods, detection algorithms, diagnostic models, and evaluation metrics. Emphasis is placed on the role of machine learning, statistical inference, and trace-based methods in detecting and localizing faults. Furthermore, the survey highlights the limitations of existing techniques in terms of scalability, explainability, and real-time applicability. By consolidating scattered research into a unified analytical framework, this work contributes a detailed taxonomy of methods and a critical discussion of their effectiveness in container-based virtualized environments. The survey provides a representative and timely overview of the field, helping to map current progress and highlight remaining gaps. Finally, it outlines promising directions such as hybrid detection architectures, automated trace reasoning, and context-aware root-cause localization, offering guidance for researchers and practitioners seeking to enhance reliability, observability, and operational resilience in microservice systems.
Overview
CoChat Summary
TL;DR
This survey systematically reviews anomaly detection and root-cause identification (RCI) in microservice architectures, organizing the literature by data source, detection method, and RCI technique. It compares reported performance across methods and highlights trustworthiness, testbeds, datasets, and open challenges for distributed diagnosis.
Key findings
- The survey covers data collection via logs, distributed traces, and monitoring metrics; anomaly detection via unsupervised, supervised, semi-supervised, reinforcement, trace-comparison, and statistical methods; and RCI via machine learning, graph-based, and statistical techniques.
- For anomaly detection, statistical methods had the highest accuracy (99.2%), while trace comparison had the highest recall (99.0%), precision (92.0%), and F1-score (98.2%); supervised learning had high precision/F1/accuracy but lower recall, and unsupervised learning had high precision but low accuracy and recall.
- For RCI, machine learning was most promising (precision 94.9%, recall 98.0%, F1 99.0%, accuracy 94.3%), followed by graph-based methods (precision 92.7%, recall 89.7%); statistical methods had precision 85.0%, recall 88.0%, F1 85.8%, and accuracy 99.0%.
- Top individual anomaly detection methods included tree-matching (P 97%, R 99%, F1 98%), deep variational Bayesian networks (P 98%, R 97%, F1 97%), SVM (F1 96.4%), VGAE+LSTM-AE (P 97.6%, R 93%, F1 95%), and GAT (P 90.5%, R 98.7%, F1 94.4%); top RCI methods included MLP (P 99.8%, R 98%, F1 99%), metric causality graph (P 97%), RPCA (P 90%), SBFL (P 82%, R 93%, F1 87%), and PRV (P 86%, R 83%, F1 84.5%).
- Train Ticket was the most commonly used open-source testbed, followed by Sock Shop, PyMicro, and Death Star; AIOps Challenge 2020/2021 datasets were the most cited (nine publications), and proprietary platforms included Alibaba, Netflix, IBM BlueMix, and eBay. The survey also proposes trust dimensions for distributed diagnosis: reliability, explainability, consistency, robustness, causal validity, scalability, and telemetry integrity.
Methods
- PRISMA-based systematic literature review searching seven digital libraries (IEEE Xplore, ACM Digital Library, Science Direct, Wiley Online Library, Springer Link, Scopus, Web of Science), limited to English publications after 2012, with search conducted December 2021–May 2025.
- The selection process reports 10,485 initial records and 143 included publications (141 via PRISMA plus 2 from other methods); an earlier sentence states 1,700 results and 117 included publications.
- Studies were categorized by data collection method, anomaly detection method, RCI method, fault type, and performance metrics; performance comparison selected methods with more than 80% values in the most frequent metrics (accuracy, F1-score, precision, recall).
Limitations
- The paper states: scope limited to English studies from 2012 to May 2025; possible selection bias despite seven libraries; direct performance comparisons should be interpreted cautiously due to heterogeneous datasets, experimental setups, and metrics; and rapidly evolving field means newer methods beyond the cut-off are not captured.
- The text makes evident inconsistent publication counts (1,700/117 vs. 10,485/143) and frequent missing performance values (“n.a.”), which complicate direct method comparison.
- The paper also notes that no datasets were generated or analysed during the current study, and that many surveyed works use different testbeds and datasets, limiting comparability.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures
![Fig. 1 Microservice vs. Monolithic architectures (adapted from [38] and [39])](https://figures.cochat.ai/paper-figures/a1ec89a530804676a5e0a1fe67ef3e10/anomaly-detection-and-root-cause-identification-in-microservices-a-survey-figure-1.png)
![Fig. 2 Microservice anomaly classification tree (adapted from [ 44])](https://figures.cochat.ai/paper-figures/cdb4d2d23f3743ca938927b6e20041c3/anomaly-detection-and-root-cause-identification-in-microservices-a-survey-figure-2.png)















Verification
Claims the paper makes
32 Valid · 35 Not valid · 109 Not checked · 118 Own results and interpretations
At least one cited source doesn’t support this claim or points to a different paper, and none confirmed it — see the verdicts below.
In the paper
“MicroHECL [7], MicroNet [132], and Diag-Fusion [107] apply the 3-sigma rule to detect extreme deviations in traces, metrics, and invocation latencies.”
§ Anomaly detection and root-cause identification in microservices: a survey > 4.4.2 Performance anomalies
What the sources say
- Disputedcites [7] Liu, D., He, C., Peng, X., Lin, F., Zhang, C., Gong, S., Li, Z., Ou, J., Wu, Z.: MicroHECL: High-efficient root cause localization in large-scale microservice systems. In: Proceedings of the 2021 IEEE/ACM 43rd International Conference on Software Engineering: Software Engineering in Practice (ICSE-SEIP), pp. 338–347 (2021). https://doi.org/10.1109/ICSE-SEIP52600.2021.00043 sourceExperimental — accuracy not yet measured
The Localization Accuracy discussion (p. 7) attributes use of the 3-sigma rule to Microscope, not MicroHECL. It explicitly contrasts MicroHECL’s anomaly-specific models with that rule. The separate claims about MicroNet and Diag-Fusion are attached to other citations and are not judged here. (LLM, conf=0.90)
Checked by a reasoning model — may be wrong.
- Not checked — no full text of the sourcecites [132] Yang, J., Guo, Y., Chen, Y., Zhao, Y.: MicroNet: Operation aware root cause identification of microservice system anomalies. IEEE Transactions on Network and Service Management (2024). https://doi.org/10.1109/TNSM.2024.3387552 source
- Not checked — no full text of the sourcecites [107] Zhang, S., Jin, P., Lin, Z., Sun, Y., Zhang, B., Xia, S., Li, Z., Zhong, Z., Ma, M., Jin, W., Zhang, D., Zhu, Z., Pei, D.: Robust failure diagnosis of microservice system through multimodal data. IEEE Transactions on Services Computing 16(6), 3851–3864 (2023). https://doi.org/10.1109/TSC.2023.3290018 source
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Disputed
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Although outlier detection has been studied since the 19th century [12], there is still no universal solution for modern distributed systems [13].”
§ Anomaly detection and root-cause identification in microservices: a survey > 1 Introduction
What the sources say
- Partly supportedcites [13] Pang, G., Shen, C., Cao, L., Hengel, A.V.D.: Deep learning for anomaly detection: A review. ACM Comput. Surv. 54(2), 38 (2021). https://doi.org/10.1145/3439950 sourceExperimental — accuracy not yet measured
Section 2 (p. 2) describes unsolved anomaly-detection challenges, and Section 2.2 (p. 3) says current methods still often produce false positives. This supports the general point that anomaly detection remains difficult, but the paper does not establish the more specific claim that there is no universal solution for modern distributed systems. (LLM, conf=0.84)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“As services may scale from a few to thousands of instances across dynamic infrastructures, even minor faults such as misconfiguration, hardware failure, or queue overload can propagate rapidly along service call chains [7,20].”
§ Anomaly detection and root-cause identification in microservices: a survey > 1 Introduction
What the sources say
- Partly supportedcites [7] Liu, D., He, C., Peng, X., Lin, F., Zhang, C., Gong, S., Li, Z., Ou, J., Wu, Z.: MicroHECL: High-efficient root cause localization in large-scale microservice systems. In: Proceedings of the 2021 IEEE/ACM 43rd International Conference on Software Engineering: Software Engineering in Practice (ICSE-SEIP), pp. 338–347 (2021). https://doi.org/10.1109/ICSE-SEIP52600.2021.00043 sourceExperimental — accuracy not yet measured
The Introduction (p. 1) says that services can have several to thousands of instances and that any anomaly in service quality may propagate along service call chains. It does not identify misconfiguration, hardware failure, or queue overload as examples, or say that propagation is rapid. (LLM, conf=0.86)
- Partly supportedcites [20] Islam, M.S., Pourmajidi, W., Zhang, L., Steinbacher, J., Erwin, T., Miranskyy, A.: Anomaly detection in a large-scale cloud platform. In: Proceedings of the 43rd International Conference on Software Engineering: Software Engineering in Practice. ICSE-SEIP ’21, pp. 150–159. IEEE Press, Virtual Event, Spain (2021). https://doi.org/10.1109/ICSE-SEIP52600.2021.00024 sourceExperimental — accuracy not yet measured
The Introduction (p. 2) names misconfiguration, hardware failure, and processing-queue saturation as sources of trouble, and says a problematic component can affect other inter-related components and clients. It does not establish that even minor faults propagate rapidly or trace their spread along service call chains. (LLM, conf=0.86)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“Only two survey publications [26,27] directly related to our work were found in the most relevant digital libraries (IEEE Xplore, ACM Digital Library, Science Direct, Wiley Online Library, Springer Link, Scopus, and Web of Science).”
§ Anomaly detection and root-cause identification in microservices: a survey > 2 Related surveys and contribution
What the sources say
- Supportedcites [26] Soldani, J., Brogi, A.: Anomaly detection and failure root cause analysis in (micro) service-based cloud applications: A survey. ACM Comput. Surv. 55(3) (2022) https://doi.org/10.1145/3501297 source
The citation identifies this paper as one of the survey publications the citing authors found. In Related Work (p. 29), the paper explicitly describes itself as a survey of anomaly detection and root-cause analysis techniques for multi-service applications. The count of two and the list of libraries describe the citing authors’ search, not a finding attributed to this source. (LLM, conf=0.88)
- Not checked — no full text of the sourcecites [27] Zhang, S., Xia, S., Fan, W., Shi, B., Xiong, X., Zhong, Z., Ma, M., Sun, Y., Pei, D.: Failure diagnosis in microservice systems: A comprehensive survey and analysis. ACM Trans. Softw. Eng. Methodol. (2025). https://doi.org/10.1145/3715005 source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“large-scale platforms, anomaly detectors integrated directly with existing log infrastructures emphasize scalability and elasticity, adapting to dynamic and serverless architectures [20]”
§ Anomaly detection and root-cause identification in microservices: a survey > 4.2.3 Monitoring based (metric based)
What the sources say
- Supportedcites [20] Islam, M.S., Pourmajidi, W., Zhang, L., Steinbacher, J., Erwin, T., Miranskyy, A.: Anomaly detection in a large-scale cloud platform. In: Proceedings of the 43rd International Conference on Software Engineering: Software Engineering in Practice. ICSE-SEIP ’21, pp. 150–159. IEEE Press, Virtual Event, Spain (2021). https://doi.org/10.1109/ICSE-SEIP52600.2021.00024 source
Section 2.1 (p. 3) describes using the platform’s existing message-queue log pipeline, and Section 5 (p. 12) says the deployed detector obtains logs from that queue. Section 3.1 (p. 5) identifies scalability and elasticity as design goals; Section 6.2.1 (p. 13) explains that its serverless architecture autoscales as log-arrival patterns change. (LLM, conf=0.91)
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
71 more claims · 9 with their cited sources checked
Create a free account to see all 294 claims and every source verdict.
- Probably not a substantial claim. 152
- It may state two separate claims as one. 120
- The quote only partly states this claim. 2
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Anomaly detection and root-cause identification in microservices: a survey > Anomaly Detection Method b
“Chen et al. [104] 2022 M SL P F1: 82.5”
§ Anomaly detection and root-cause identification in microservices: a survey > 5.4 RQ4: What is the performance of the various identified methods?
“Chen et al. [104] applied Adaptive Pattern Learning, but provided only precision (82.5%).”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Anomaly detection and root-cause identification in microservices: a survey > 2 Related surveys and contribution
“reviewing 1,700 results”
§ Anomaly detection and root-cause identification in microservices: a survey > 4.1.4 Selection process
“The process started with seven initial queries, as described in Figure 3, from which 10485 publications were identified; thus, the process of finding duplicates was initiated.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Anomaly detection and root-cause identification in microservices: a survey > 4.7 Methods comparison
“The Statistical methods showed acceptable performance, with a precision of 85.0%, recall of 88.0%, and an F1-Score of 85,8%.”
§ Anomaly detection and root-cause identification in microservices: a survey > 5.3 RQ3: How can the causes of these anomalies be identified?
“the statistical method performs reasonably well with high accuracy but lacks comprehensive assessment without F1-Score data.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Anomaly detection and root-cause identification in microservices: a survey > Anomaly Detection Method b
“Ghorbani et al. [111] 2021 M UL S FPR: 0.01, P: +9, A: +14”
§ Anomaly detection and root-cause identification in microservices: a survey > 5.3 RQ3: How can the causes of these anomalies be identified?
“Ghorbani et al. [111] applied Autoencoders and achieved a Precision increase of +9% and a Recall increase of +14%.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Anomaly detection and root-cause identification in microservices: a survey > 3.8 Metrics for performance evaluation of anomaly classification
“Accuracy is calculated by comparing all correct predictions, including both true and false positives, to the total number of predictions made by the model.”
§ Anomaly detection and root-cause identification in microservices: a survey > 7 Conclusion
“Accuracy = TP + TN TP + FP + TN + FN”
Citations
Cites
Chandola, V., Banerjee, A., & Kumar, V. (2009) · ACM Computing Surveys · doi:10.1145/1541880.1541882 · Cited by 11,687 · cited 4×
“1 Microservice vs. Monolithic architectures (adapted from [38] and [39])”
InfluentialCausality: Models, Reasoning, and Inference
Pearl, J. (2009) · doi:10.1017/cbo9780511803161 · Cited by 11,653
Arrieta, A. B., Díaz-Rodríguez, N., Del Ser, J., Bennetot, A., Tabik, S., Barbado, A., García, S., Gil-López, S., Molina, D., Benjamins, R., et al. (2019) · arXiv · doi:10.48550/arxiv.1910.10045 · Cited by 11,321
Barredo Arrieta, A., Díaz-Rodríguez, N., Del Ser, J., Bennetot, A., Tabik, S., Barbado, A., Garcia, S., Gil-Lopez, S., Molina, D., Benjamins, R., et al. (2020) · Information Fusion · doi:10.1016/j.inffus.2019.12.012 · Cited by 10,423
Edge computing: Vision and challenges
Shi, W., Cao, J., Zhang, Q., Li, Y., & Xu, L. (2016) · IEEE Internet of Things Journal · doi:10.1109/jiot.2016.2579198 · Cited by 8,733 · cited 1×
“For anomaly detection in microservices, TDAI is particularly relevant in edge and IoT scenarios, where embedded AI models are deployed close to the data source [237, 238].”
Federated learning: Challenges, methods, and future directions
Li, T., Sahu, A. K., Talwalkar, A., & Smith, V. (2020) · IEEE Signal Processing Magazine · doi:10.1109/msp.2020.2975749 · Cited by 6,210 · cited 1×
“Trusted Distributed AI extends the principles of federated and collaborative learning with trust guarantees, ensuring that distributed models can be trained and deployed without exposing sensitive data or compromising model integ - rity [235, 236].”
Practical Secure Aggregation for Privacy-Preserving Machine Learning
Bonawitz, K., Ivanov, V., Kreuter, B., Marcedone, A., McMahan, H. B., Patel, S., Ramage, D., Segal, A., & Seth, K. (2017) · Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security · doi:10.1145/3133956.3133982 · Cited by 3,956 · cited 1×
“By combining local inference with trusted aggregation, TDAI can enhance both privacy and robustness of anomaly detection pipelines [239].”
A Survey of Outlier Detection Methodologies
Hodge, V., & Austin, J. (2004) · Artificial Intelligence Review · doi:10.1007/s10462-004-4304-y · Cited by 3,596
Towards A Rigorous Science of Interpretable Machine Learning
Doshi-Velez, F., & Kim, B. (2017) · arXiv · doi:10.48550/arxiv.1702.08608 · Cited by 3,121
Deep Learning for Anomaly Detection: A Review
Pang, G., Shen, C., Cao, L., & Hengel, A. V. D. (2022) · ACM Computing Surveys · doi:10.1145/3439950 · Cited by 2,837
Research works printed but not linked: 5
- Yuan, D., Park, S., Zhou, Y., Savage, S.: Sherlock: Performance bug diagnosis in production systems. In: Proceedings of the 10th USENIX Symposium on Operating Systems Design and Implementation (OSDI 2012), pp. 1–14. USENIX Association, Berkeley, CA, USA (2012)
- Pahl, M.-O., Aubet, F.-X.: All Eyes on You: Distributed Multi-Dimensional IoT Microservice Anomaly Detection. In: Proceedings of the 2018 14th International Conference on Network and Service Management (CNSM), pp. 72–80 (2018). ISSN: 2165-963X
- Sutton, R.S., Barto, A.G.: Reinforcement Learning, Second Edition: An Introduction. Adaptive Computation and Machine Learning series. MIT Press. https://books.google.pt/books?id=uWV0DwAAQBAJ
- Harnik, D., Margalit, O., Naor, M., Tsfadia, A.: Confidential logging using trusted execution environments. In: Proceedings of the 28th USENIX Security Symposium, pp. 1627–1644 (2019)
- Glymour, M., Glymour, C., Jewell, N.: Causal Inference in Statistics: A Primer. Wiley, Hoboken, NJ, USA (2016)
Cited by
Awartani, A. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7583047
Metadata
- Authors
- Barata, L. M., Sequeira, S., Lopes, E., Inácio, P. R. M., & Freire, M. M.
- Affiliations
- Universidade Politécnica de Castelo Branco (PT)
- University of Beira Interior (PT)
- Corresponding author
- Barata, Luís M.
- Venue
- Cluster Computing
- Published
- 2026-06
- Volume
- 29
- Issue
- 5
- ISSN
- 1386-7857 (ISSN-L)
- 1573-7543
- OpenAlex
- W7163357751
- Cited by
- 4
- References
- 192
- Citations by year
- 2026: 4
- FWCI
- 9.49
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- creativecommons.org/licenses/by/4.0 (tdm)
- creativecommons.org/licenses/by/4.0 (vor, 2 d)
- Locations
- Cluster Computing (publishedVersion, cc-by)
- Funders
- Universidade da Beira Interior
- Topics
- Physical Sciences › Computer Science › Computer Networks and Communications › Software System Performance and Reliability
- Physical Sciences › Computer Science › Information Systems › Cloud Computing and Resource Management
- Physical Sciences › Computer Science › Computer Networks and Communications › Distributed systems and fault tolerance
- Keywords
- microservices, anomaly detection, root cause identification, fault location, system monitoring, distributed systems, machine learning for fault detection, statistical inference, container-based virtualization, observability, operational resilience
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-05, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-05 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Anomaly detection and root-cause identification in microservices: a survey. https://app.cochat.ai/papers/10.1007/s10586-026-06095-9
- The paper
- Barata, L. M., Sequeira, S., Lopes, E., Inácio, P. R. M., & Freire, M. M. (2026). Anomaly detection and root-cause identification in microservices: a survey. Cluster Computing. https://doi.org/10.1007/s10586-026-06095-9
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.