Bachelorarbeit BCLR-2020-34

Farley, Andreas: Continuous Performance Testing of FaaS and Microservices based on TOSCA Topology and Orchestration Specifications.
Universität Stuttgart, Fakultät Informatik, Elektrotechnik und Informationstechnik, Bachelorarbeit Nr. 34 (2020).
91 Seiten, englisch.

With the adoption of cloud computing growing steadily, enterprises are fighting with cloud cost optimization, as their applications are sluggish in regards to the interchangeability of individual components. Additionally, deployment and provisioning of cloud applications requires multiple individual steps, which are often scripted manually. To combat these issues, Topology and Orchestration Specification for Cloud Applications (TOSCA), a language to describe cloud applications in all regards, including management operations, was developed. Rational decomposition and orchestration for serverless computing (RADON), a proposed TOSCA based framework for microservices and Function-as-a-Service (FaaS), aims to support Development and Operations (DevOps) practices. One part of these practices is continuous testing, which prompted the question, how to generate tests from a TOSCA service template and execute them afterwards. To answer this question, the TOSCA meta model was analyzed to find an approach in regards to test generation and execution. We formulated an approach and implemented it in a Continuous Testing Tool (CTT) called RadonCTT. We then evaluated RadonCTT using a demonstration cloud application called Toy-Example. This was done to see whether or not our test generation and execution was functional. Additionally, we evaluated how accurate the test results were. Finally, we evaluated if RadonCTT could be used to cover use cases other than performance testing. Our evaluation shows that our approach can be implemented and that the implementation can be used to generate and execute performance tests, which produce accurate results. Additionally, our results show that RadonCTT supports use cases other than performance testing. We therefore conclude that our approach is a feasible solution to automatic test generation and execution from a TOSCA service template. Our evaluation does not, however, show that our implementation produces accurate results for every cloud application. This is due to a lack of applications available for our evaluation.

Abteilung(en)Universität Stuttgart, Institut für Softwaretechnologie, Sichere und Zuverlässige Softwaresysteme
Betreuervan Hoorn, Dr. Andre; Düllmann, Thomas F.; Okanovic, Dusan
Eingabedatum10. November 2020
   Publ. Informatik