ANOMALY DETECTION ON KUBERNETES

• Main Author: Supendi, Kevin
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/28445
• Institution: Institut Teknologi Bandung

Time-series anomaly detection is an interesting topic especially for monitoring. Every monitoring system could benefit from anomaly detection. Earlier detection could prevent problems and improve confidence for monitoring system. <br /> <br /> <br /> Kubernetes is a popular container orchestration technology, but it lacks anomaly detection features that is tested and open source. In this thesis, anomaly detection solution for Kubernetes was created using Hierarchical Temporal Memory algorithm. Then the performance between single time-series anomaly detection and multiple time-series detection was compared. <br /> <br /> <br /> The thesis concludes that the proposed solution could work well in Kubernetes environment, with moderate performance. Test result shows that single time-series anomaly detection is better than multiple time-series anomaly detection, because multiple time-series detection needs more specific test cases. <br />