Master Thesis MSTR-2017-23

BibliographyDüllmann, Thomas F.: Performance anomaly detection in microservice architectures under continuous change.
University of Stuttgart, Faculty of Computer Science, Electrical Engineering, and Information Technology, Master Thesis No. 23 (2017).
89 pages, english.
Abstract

The idea of DevOps and agile approaches like Continuous Integration (CI) and microservice architectures are bocoming more and more popular as the demand for flexible and scalable solutions is increasing. By raising the degree of automation and distribution new challenges in terms of application performance monitoring arise because microservices are possibly short-lived and may be replaced within seconds. The fact that microservices are added and removed on a regular basis brings new requirements in the way anomaly detection is conducted as these changes could also be the cause for anomalies. This work proposes to take information about such events into account to improve the anomaly detection quality. Additionally, a meta model for microservice environments and supplemental tooling was developed that can be used to generate actual microservice environments from an instance of such a meta model. The generation tool also comprises the means to instrument the generated microservices with Kieker to collect monitoring data and generate supplemental files to be able to create Docker images and Kubernetes configuration files which allow to run the microservices on a Kubernetes cluster. In the evaluation such a generated microservice environment is run in a lab experiment with delay injections. The obtained data were then used to evaluate a customized version of the RanCorr approach which was modified to fit in the microservice context and a newly developed basic approach named Event-aware Anomaly Revisor (EAR) that takes event information into account when conducting anomaly detection. The evaluation results showed, that the customized RanCorr approach could not satisfy the expectations in terms of improvement of the anomaly detection results while the EAR approach could slightly improve the anomaly detection quality.

Full text and
other links
Volltext
Department(s)University of Stuttgart, Institute of Software Technology, Software Reliability and Security
Superviser(s)van Hoorn, Dr. André
Entry dateMay 28, 2019
   Publ. Computer Science