The DOI matches its Crossref record; 18 of the 18 references it cites were found in a registry or an index; 5 of the 18 sources its claims cite were checked for support.
A Survey of Anomaly Detection and Root Cause Analysis Techniques for Artificial Intelligence for IT Operations (AIOps) in Large-Scale Software Systems
Open accessAbstract
With the widespread adoption of cloud computing and microservice architectures, software system complexity has grown exponentially, making traditional operation and maintenance (O&M) paradigms relying on predefined rules and manual expertise unsustainable. Artificial Intelligence for IT Operations (AIOps) has thus emerged as a promising solution [1]. This review focuses on mainstream AIOps practice for large-scale cloud-native software systems from 2020 to 2025. Unlike prior surveys centered on either multimodal data fusion or LLM-based AIOps, it integrates multimodal observability data, anomaly detection and root cause analysis methods, and recent OpsLLM applications into a unified framework. This paper systematically sorts out core AIOps technologies, including multimodal data analysis based on logs, metrics, and traces, the evolution of anomaly detection from statistical methods to deep learning, and root cause localization algorithms based on topological graphs and causal inference. It also summarizes typical application scenarios of operation-specific LLMs (OpsLLM) from 2023 to 2025. Finally, it identifies core challenges of current AIOps, such as data annotation bottlenecks and insufficient model interpretability, and outlooks future development trends.
Overview
CoChat Summary
TL;DR
This survey reviews AIOps techniques for large-scale cloud-native software systems, integrating multimodal observability data (logs, metrics, traces), anomaly detection, root cause analysis, and OpsLLM applications into a unified framework. It covers mainstream practice from 2020–2025 and identifies data annotation bottlenecks, insufficient model interpretability, and computing/real-time engineering constraints as core open challenges.
Key findings
- Anomaly detection is evolving from shallow statistical methods for single indicators toward deep semantic correlation analysis of multi-source data; the paper's comparison table rates statistical methods (3-Sigma, ARIMA) as "Medium" accuracy but "Very High" efficiency and interpretability with no label requirement, supervised learning (SVM, BiLSTM) as "High" accuracy but "High" label requirement, unsupervised reconstruction (Donut, VAE) as "Relatively High" accuracy with low interpretability, and semantic pre-training (LogBERT) as "Very High" accuracy but "Relatively Low" efficiency.
- RCA is shifting from explicit topology-based correlation search to cross-modal semantic logical reasoning: trace-based methods (MicroRCA, CloudRCA) are mature but only locate faulty nodes without explaining causes; causal graph methods (PC algorithm and variants) are mathematically interpretable but suffer the curse of dimensionality, data noise, and inconsistent sampling; cross-modal fusion (TraceAnomaly, CORAL) offers full-stack diagnosis but faces clock alignment, feature-weight allocation, high computational complexity, and strict training-data requirements.
- Since 2023, LLMs (e.g., GPT-4) introduced a semantic reasoning paradigm; OpsLLM models such as Owl and XrayGLM offer strong zero/few-shot generalization and overcome single-task retraining, but face high inference costs and hallucination risks, with large-scale real-time economic feasibility unverified.
- The paper identifies three core industrial challenges: scarcity of high-quality annotated fault data, insufficient model interpretability, and engineering bottlenecks in computing power and real-time response for large-scale scenarios.
Methods
- Design: a systematic literature review/survey of AIOps for large-scale cloud-native software systems, organized around a unified analytical framework spanning multimodal data, anomaly detection, RCA, and OpsLLM.
- Scope: foundational methods before 2020 are cited mainly to explain technical origins, while system practices and LLM-driven applications are centered on 2020–2025; the O&M context is cloud-native, containerized, and microservice-based services in public/private clouds and datacenters, with isolated device-level maintenance excluded.
- Analysis: qualitative synthesis and comparative tabulation of method families (Table I compares logs, metrics, and traces; Table II compares anomaly detection method categories across accuracy, efficiency, interpretability, and label requirement), plus presentation of representative formulas (3-Sigma, Random Walk with Restart anomaly score, OpsLLM SFT cross-entropy loss).
Limitations
- Stated: the paper notes that current AIOps faces data annotation bottlenecks, insufficient model interpretability, and engineering bottlenecks in computing power and real-time response; it also notes that LLM-based log diagnosis is constrained by context windows and that large-scale real-time economic feasibility is unverified.
- Evident from the text: as a survey, it reports no primary experiments, datasets, or quantitative evaluation of its own; the comparative ratings in Tables I and II are qualitative labels without reported benchmarks or metrics; and the scope is restricted to cloud-native/microservice environments, with edge and isolated device-level maintenance explicitly outside the main scope.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
3 Valid · 8 Not valid · 9 Not checked · 14 Own results and interpretations
A claim is checked only against the cited paper’s full text, never its abstract, and none is available for its sources.
In the paper
“Its paradigm has shifted from traditional rule matching to deep Natural Language Processing (NLP) models, represented by DeepLog [12] and LogRobust [13] .”
What the sources say
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.
In the paper
“LLMs can leverage the general knowledge acquired during pretraining [21] to perform logical reasoning in combination with operational context.”
What the sources say
- Not matchedcites [21] Gao, 2024, A Survey on Large Language Models for AIOps: Progress, Forecasting and Challenges, arXiv preprint
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
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.
- Not matched
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Driven by LLMs, AIOps has expanded from single alert monitoring to interactive fault troubleshooting, with two core scenarios.”
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.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“In summary, RCA technologies are evolving from explicit topology-based correlation search to cross-modal semantic logical reasoning.”
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.
In the paper
“Leveraging machine learning and Natural Language Processing (NLP) algorithms, AIOps can autonomously learn normal baselines from massive historical data.”
What the sources say
- Not checked — no full text of the sourcecites [1] 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.
3 more claims
Create a free account to see all 34 claims and every source verdict.
- Probably not a substantial claim. 13
- It may state two separate claims as one. 18
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
§ A. Trace-Based Localization Methods
“Typical algorithms (MicroRCA [15] , CloudRCA [16] )”
§ REFERENCES
“- [15] P. Wang et al., "CloudRCA: A Root Cause Analysis Framework for Cloud Computing Platforms," in Proc. 27th ACM Int. Conf. Inf. Knowl. Manag. (CIKM), Torino, Italy, 2018, pp. 1153-1162. - [16] L. Wu et al., "MicroRCA: Root Cause Localization of Performance Issues in Microservices," in Proc. IEEE/IFIP Netw. Oper. Manag. Symp. (NOMS), Bordeaux, France, 2021, pp. 1-7.”
Citations
Cites
DeepLog: Anomaly Detection and Diagnosis from System Logs through Deep Learning
Du, M., Li, F., Zheng, G., & Srikumar, V. (2017) · Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security · doi:10.1145/3133956.3134015 · Cited by 1,835
Robust Anomaly Detection for Multivariate Time Series through Stochastic Recurrent Neural Network
Su, Y., Zhao, Y., Niu, C., Liu, R., Sun, W., & Pei, D. (2019) · Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining · doi:10.1145/3292500.3330672 · Cited by 1,804 · cited 1×
“It is divided into two major technical routes: reconstruction error-based and prediction deviation-based models, with the VAE-based Donut [7] , LSTM and its variants [8] as representative algorithms.”
Drain: an online log parsing approach with fixed depth tree
He, P., Zhu, J., Zheng, Z., & Lyu, M. R. (2017) · 2017 IEEE International Conference on Web Services (ICWS) · doi:10.1109/icws.2017.13 · Cited by 943 · cited 1×
“The industry widely adopts online clustering algorithms based on parse trees such as Drain [9] , which can compress data dimensions efficiently but has insufficient discrimination for templates with high semantic similarity.”
Unsupervised Anomaly Detection via Variational Auto-Encoder for Seasonal KPIs in Web Applications
Xu, H., Feng, Y., Chen, J., Wang, Z., Qiao, H., Chen, W., Zhao, N., Li, Z., Bu, J., Li, Z., et al. (2018) · Proceedings of the 2018 World Wide Web Conference on World Wide Web - WWW '18 · doi:10.1145/3178876.3185996 · Cited by 808
Robust log-based anomaly detection on unstable log data
Zhang, X., Xu, Y., Lin, Q., Qiao, B., Zhang, H., Dang, Y., Xie, C., Yang, X., Cheng, Q., Li, Z., et al. (2019) · Proceedings of the 2019 27th ACM Joint Meeting on European Software Engineering Conference and Symposium on the Foundations of Software Engineering · doi:10.1145/3338906.3338931 · Cited by 761
LogAnomaly: Unsupervised Detection of Sequential and Quantitative Anomalies in Unstructured Logs
Meng, W., Liu, Y., Zhu, Y., Zhang, S., Pei, D., Liu, Y., Chen, Y., Zhang, R., Tao, S., Sun, P., et al. (2019) · Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2019/658 · Cited by 708
Time-Series Anomaly Detection Service at Microsoft
Ren, H., Xu, B., Wang, Y., Yi, C., Huang, C., Kou, X., Xing, T., Yang, M., Tong, J., & Zhang, Q. (2019) · Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining · doi:10.1145/3292500.3330680 · Cited by 623
LogBERT: Log Anomaly Detection via BERT
Guo, H., Yuan, S., & Wu, X. (2021) · 2021 International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn52387.2021.9534113 · Cited by 454
MicroRCA: Root Cause Localization of Performance Issues in Microservices
Wu, L., Tordsson, J., Elmroth, E., & Kao, O. (2020) · NOMS 2020 - 2020 IEEE/IFIP Network Operations and Management Symposium · doi:10.1109/noms47738.2020.9110353 · Cited by 265
Unsupervised Detection of Microservice Trace Anomalies through Service-Level Deep Bayesian Networks
Liu, P., Xu, H., Ouyang, Q., Jiao, R., Chen, Z., Zhang, S., Yang, J., Mo, L., Zeng, J., Xue, W., et al. (2020) · 2020 IEEE 31st International Symposium on Software Reliability Engineering (ISSRE) · doi:10.1109/issre5003.2020.00014 · Cited by 252
Research works printed but not linked: 8
- Cambria, 2023, Multimodal Data Fusion for AIOps: A Survey on Architectures and Algorithms, IEEE Transactions on Network and Service Management, 20, 1567
- Han, 2023, A Closed-Loop Architecture for Intelligent Operation and Maintenance in Cloud-Native Systems, IEEE Internet of Things Journal, 10, 9521
- Li, 2018, Robust Anomaly Detection for Multivariate Time Series Through GAN, in Proc. 24th ACM SIGKDD Int. Conf. Knowl. Discov. Data Min. (KDD), 1516
- Wang, 2018, CloudRCA: A Root Cause Analysis Framework for Cloud Computing Platforms, in Proc. 27th ACM Int. Conf. Inf. Knowl. Manag. (CIKM), 1153
- Chen, 2022, Practical Root Cause Analysis via Causal Learning and Network Flows, IEEE Transactions on Network and Service Management, 19, 4818
- Zhang, 2022, Pulling Out All the Stops: Hierarchical and Multi-Scale Anomaly Detection, IEEE Transactions on Reliability, 71, 1162
- Gao, 2024, A Survey on Large Language Models for AIOps: Progress, Forecasting and Challenges, arXiv preprint
- Sun, 2024, Trace-Enabled Causal Analysis with LLMs, in Proc. 46th Int. Conf. Softw. Eng. (ICSE), 185
Metadata
- Authors
- Yu, E.
- Affiliations
- Xi'an Jiaotong University (CN)
- Corresponding author
- Yu, Erzhan
- Venue
- 2026 6th International Conference on Artificial Intelligence and Industrial Technology Applications (AIITA)
- Publisher
- IEEE
- Published
- 2026-04-10
- Pages
- 391-395
- OpenAlex
- W7165797161
- Semantic Scholar
- 8a323aa295f267aeb09387b7e7e4635c4d960052
- Cited by
- 0
- References
- 13
- FWCI
- 0.00
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Locations
- International Conference on Artificial Intelligence and Industrial Technology Applications (AIITA) (publishedVersion, cc-by)
- 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 › Information Systems › Software Engineering Research
- Keywords
- AIOps, anomaly detection, root cause analysis, large software systems, cloud-native systems, microservices, deep learning anomaly detection, topological graphs, causal inference, model interpretability, data annotation
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 2026-10-11, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-11 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
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: A Survey of Anomaly Detection and Root Cause Analysis Techniques for Artificial Intelligence for IT Operations (AIOps) in Large-Scale Software Systems. https://app.cochat.ai/papers/10.1109/aiita69518.2026.11567316
- The paper
- Yu, E. (2026). A Survey of Anomaly Detection and Root Cause Analysis Techniques for Artificial Intelligence for IT Operations (AIOps) in Large-Scale Software Systems. 2026 6th International Conference on Artificial Intelligence and Industrial Technology Applications (AIITA). https://doi.org/10.1109/aiita69518.2026.11567316
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.