Dissertation, TU-Vienna, 2010

 

Modern IT systems are evolving from monolithic software packages to distributed and highly interactive integration solutions. Concurrent systems and system components thereby need to be coordinated on many different levels, including the partitioning of processing units in multi-core processor systems, the coordination of workflows in enterprise systems, and the global mashup of software services in the Internet. As a direct consequence, the development of novel coordination mechanisms that can meet the strong requirements of such highly interactive systems is becoming a critical factor.

 

With the goal to develop such new coordination mechanisms, several research projects investigated how space-based coordination models can be integrated with knowledge representation technologies from the Semantic Web field. It was shown that extending the space-based interaction paradigm with mechanisms for the representation and automated evaluation of information and its relationships has various advantages for developing complex coordination solutions. However, as these so-called "semantic spaces" represent a comparably young research field, no standardized solutions have been established yet and the available implementations often differ in terms of performance and scope of functionality. Hence, it is often not clear how to conceive coordination solutions with semantic spaces and which are the strengths and weaknesses of such solutions.

 

In the course of this thesis, it is investigated what characterizes coordination solutions with semantic spaces and which advantages and disadvantages are entailed by the integration with knowledge representation technologies. The thesis further addresses frequently observed performance problems and presents an optimization mechanism that allows for the efficient detection of changes of the knowledge represented in a semantic space. To ensure that the employment of such optimization mechanisms does not lead to undesired side-effects, a formal method for the specification of such systems is developed. This method can be used to formally describe the interface semantics and the runtime behavior of semantic spaces.

 

For the evaluation of the developed optimization mechanism, a benchmark framework for semantic spaces is implemented. The concrete performance improvements of different types of application scenarios are quantified by means of a comprehensive benchmark setting and it is analyzed which scenarios are best suited for the developed optimization mechanisms.