[Back]

S. Meixner:
"Data-Driven Automatic Deployment in Edge Computing";
Supervisor: S. Schulte, D. Schall; Institute of Information Systems Engineering, Distributed Systems Group, 2018; final examination: 2018-06-04.

With the growing popularity of the Internet of Things, we see a trend towards combiningtraditional cloud computing with resources available at the edge of the network. Thisway it becomes possible to exploit the complementary characteristics of both typesof platforms. However, unifying the two types of platforms poses new challenges todevelopers and operational staff alike, as it becomes increasingly harder to determinewhere services should run based on their non-functional- and runtime-requirements, whilesimultaneously utilizing the resources at hand in an optimal way. Manually deciding where each individual service should run, and rolling them outbecomes unfeasible, especially with a large number of individual services, which tends tobe the case in a microservice architecture. Furthermore, once the services are deployedinto production, it becomes necessary to monitor their runtime behavior to detect adeterioration of the individual services´ quality of service parameters as well as those ofthe system as a whole. Thereby, it becomes possible to take actions to prevent quality ofservice and service level agreement violations. Additionally, the collected informationcan be used to optimize future the deployment plans for the services. In this work we propose a holistic approach towards supporting developers and operationalstaff in creating and running applications that employ a microservice architecturalpattern. To realize this approach we prototypically implement a Data-Driven AutomaticDeployment framework which allows the transparent deployment of services onto cloudand edge hosts alike. Furthermore, it provides a uniform monitoring mechanism for theservices, which enables an event-based mechanism for runtime adaptation.