Service-Oriented Computing with Wireless Participants - Adaptation Mechanisms and Decision Support for Lightweight Web Service Consumption
Supervisor(s) and Committee member(s): Ralf Steinmetz (supervisor), Claudia Linnhoff-Popien (rapporteur)
At the crossroads of two hot trends of modern computer science, namely service-orientation and mobile computing, great potentials arise together with tough challenges. The list of advantages of combining these technologies is long and compelling: Outsourcing of data- and processing-intensive software tasks from mobile devices to more capable systems, quick mobile application development through the use of existing software services that are otherwise difficult to implement on mobile devices from scratch, and the list goes on. However, service-oriented messaging and description technologies are characterized by a verbose, self-descriptive nature, in order to achieve interoperability and platform-independence. This renders them often heavyweight and thus not always a good match for the resource-constrained nature of mobile, wireless devices.
In the face of the fact that very different devices will coexist in future service-oriented systems, researchers are diligently trying to tackle the aforementioned challenges by developing adaptation mechanisms. These are meant to allow the systems to dynamically and seamlessly switch to configurations that suit each particular participant. When it comes to one of the most important service-oriented technologies, namely Web services, adaptation for resource-constrained devices usually translates to an attempt to use lightweight ways of communication that are dictated by the system context.
The striving for enhancements in the field of Web service adaptation mechanisms for wireless participants starts with three simple questions: Which specific adaptation mechanisms can be used and what are their characteristics? Can the development of new adaptation mechanisms contribute to the currently existing possibilities? Which adaptation mechanisms should be used under which wireless system contexts? The systematic and scientific examination of these questions leads to the three corresponding main contributions of this thesis.
First, the conducted survey and comparison of Web service adaptation mechanisms is the first work examining the conditions of the wireless system context under which the adaptation mechanisms achieve significant benefits. Thus, the survey results can be used as a basis for examining the two further issues. Second, a new, caching-based Web service adaptation mechanism is presented. The developed mechanism is the first to enable the use of cached responses of external, i.e., third-party, Web services with guaranteed 100% freshness in an automated and generic manner. The freshness of cached objects refers to their probability of being up-to-date. The evaluation proves that the approach can lead to performance enhancements of mobile Web service invocations compared to other approaches that achieve absolute freshness. Third, provided that no single adaptation mechanism is the best-performing under all possible system contexts, the thesis offers insights with regard to the issue of corresponding decision support. Decision support algorithms that are based on the obtained survey results are developed and evaluated with focus on an important, yet uninvestigated aspect, namely the existence of missing system context data. The respective work reveals which data imputation approaches are best suited to the examined scenario.
By discussing the above issues within a well-defined future scenario of mediated Web service usage, the thesis provides further contributions, such as architectural solutions, engineering approaches, and problem formulations. All in all, its findings should be interesting for any work in the research area of mobile services.