System support for proactive adaptation

VanSyckel, Sebastian

Vorschau

PDF
Dissertation_SebastianVanSyckel.pdf - Veröffentlichte Version
Download (3MB)

URL:	https://madoc.bib.uni-mannheim.de/39016
URN:	urn:nbn:de:bsz:180-madoc-390167
Dokumenttyp:	Dissertation
Erscheinungsjahr:	2015
Ort der Veröffentlichung:	Mannheim
Hochschule:	Universität Mannheim
Gutachter:	Becker, Christian
Datum der mündl. Prüfung:	12 Juni 2015
Sprache der Veröffentlichung:	Englisch
Einrichtung:	Fakultät für Betriebswirtschaftslehre > Wirtschaftsinformatik II (Becker 2006-2021)
Fachgebiet:	004 Informatik
Normierte Schlagwörter (SWD):	Anpassung , Proaktives Computing , Ubiquitous Computing
Freie Schlagwörter (Englisch):	Proactive Adaptation , Application Adaptation , Application Configuration , Adaptation Coordination , Adaptation Control , System Support , Pervasive Computing
Abstract:	Applications in our modern, pervasive computing environments have to adapt themselves or their context in order to cope with changes. In the process, these pervasive applications should be as unobtrusive as possible, i.e., their adaptation should be automatic. In dynamic multi-user systems with shared resources and interactive applications, such adaptations cannot be scripted in advance. Instead, they have to be calculated at runtime. However, the necessary calculations quickly exceed the complexity that can be handled in real-time, i.e., without causing significant delays. The concept of proactive adaptation allows to change applications and/or context based on prediction of context and user behavior. Hence, proactive adaptation can reduce adaptation delays and avoid context interferences by determining coordinated adaptation plans ahead of time, instead of reactively when adaptation becomes necessary. Further, it helps to provide a seamless service to the user, while optimizing the overall system utility. This thesis presents a general framework and middleware-based system support for coordinated proactive adaptation in dynamic multi-user pervasive systems. The framework consists of five major components. The context interaction model and corresponding context broker offers context information, prediction, as well as actuation in a uniform fashion. The application configuration model allows applications to specify their requirements towards their context, as well as detail user preferences and duration-dependent utility and cost functions for adaptation optimization. Configuration algorithms calculate and rate all adaptation alternatives of an application given a current or predicted context and the specified rating functions, before coordination algorithms find interference-free adaptation plans for situations in which multiple applications share a context space. Finally, the adaptation control component combines the individual components of the framework in a two-dimensional control loop for proactive and fallback reactive adaptation. The prototype framework is evaluated in real-time simulations of an interactive pervasive system using recorded user traces.