Diploma and Master Theses (authored and supervised):
"Decentralized Run-time Architecture Tracking";
Supervisor: S. Dustdar, C. Mayr-Dorn;
Institute of Information Systems, Distributed Systems Group,
final examination: 2016-04-12.
Distributed applications are becoming more common, especially cloud-based applications are a major topic in recent research. These application mostly contain complex application logic and should be built for dynamic adoption. Thus monitoring of an applications components is an important topic for adaption and scalability.
There currently are different methods of adaption approaches: localized and remote.
Localized approaches do not always yield optimal solutions due to their localized views. Adoption strategies might include or effect multiple parts of the application, or might need information of those parts to create an optimal adoption strategy. Due to the localized view it is impossible to assess the impact of the adaption effects in one part of the system on other system parts, thus yielding sub-optimal adaption strategies and unforeseen side effects.
Remote approaches on the other hand provide a holistic view which allows for more intelligent adoption strategies. But existing solutions have limitations with distributed applications.
Pure probing might not be feasible (e.g. bandwidth, granularity, . . . ) and the given level of details may not be needed by the adaption control.
We propose a distributed architecture tracking framework based on architecture-based-selfadaption. By decoupling the application and the model generation itself we are able to generate a holistic view of the overall application architecture. Thus it is possible to perform optimal adaption decisions for the application.
Our proposed framework is evaluated using a simple distributed application, that is a cloudbased publish-subscribe system.
Created from the Publication Database of the Vienna University of Technology.