Institutional Repository of Vilnius University
Mašininio mokymosi modelių naudojimas Kubernetes klasterių stebėjimui ir anomalijų aptikimui /
Abstract
dc:descriptionThis thesis examines the problem of Kubernetes cluster monitoring and anomaly detection by analyzing system logs and applying machine learning methods in cloud system environments. The paper analyzed Kubernetes monitoring architecture, existing log analysis methods, and the applicability of sequence analysis models for anomaly detection. It also investigated the principles of applying long short-term memory (LSTM) neural networks and large language models (LLM) for automated Kubernetes log analysis. During the work, three open source large-scale language models were compared - Qwen, Meta-Llama and Mistral, of which Qwen achieved the best anomaly detection results. A synthetic dataset of Kubernetes access logs was also generated for anomaly detection experiments. Based on this dataset, an LSTM-based anomaly detection model was developed that can identify unusual behavior patterns and deviations from normal system operation. A NestJS-based monitoring system was developed that integrates Kubernetes API, log collection, LSTM anomaly detection model and LLM-based log interpretation system. Large language models were used to assess the pod state and generate convenient diagnostic explanations using a query methodology.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Krajeckaitė, Emilija,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD308338543&prefLang=en_US