A container-based architecture is a promising approach for building and deploying distributed systems. The expansion in terms of complexity of such infrastructures has complicated the monitoring process.
Due to the proliferation of objects and metrics to track, monitoring task face serious gaps to catch a performance drop on a single container and its effects on other dependent containers. Therefore, acting prospectively during the monitoring task by analyzing the system state helps controlling system resources and assists orchestration tools to react proactively in case of workloads anomalies. In fact, anomaly prediction is a promising option to help in tracking anomalous behavior of the overall system.
This work investigates, based on tracking resources metrics, how to predict anomalous workload behavior in a cluster architecture consisting of nodes in which different containerized application are deployed. It provides an adapted Bayesian learning approach that predicts anomalous behavior in a containerized cluster environment based on the observed metrics. Our evaluation shows a promising prediction accuracy.
| Date | 16 Jul 2021 |
|---|
| Original language | French |
|---|
| Awarding Institution | - École de technologie supérieure
|
|---|
| Supervisor | Abdelouahed Gherbi (Supervisor) |
|---|
Ben Dahsen, S. (Author),
Gherbi (Supervisor),
16 Jul 2021Student thesis: Master's thesis › Master in Engineering: Information Technology Engineering