Availability of high bandwidth networks and ubiquity of low-cost information services at the offices, homes, and lately personal mobile and wireless devices have fueled the development of new class of applications that incorporate dissemination-based data delivery approaches in addition to traditional request-response mechanisms. In this thesis, we propose a toolkit architecture that provides support to prototype and build such large-scale information dissemination systems in various application domains. We present the data structures and algorithms developed in the toolkit components which enable the use different data delivery options in a single application. Moreover, by creating a hierarchy of brokers, the toolkit can tailor information delivery to the needs of wide range of applications across various environments. The toolkit also provides facilities for performance monitoring which can allow a system developer to examine the impact of using different data delivery mechanisms.

Due to sheer volume of data available in the information dissemination systems, the use of selective dissemination mechanisms is essential in order to avoid overwhelming the users with unnecessary information. Thus, in this thesis we focus on data model and filtering algorithms in the toolkit architecture. Based on our event-based scheduler design work within the context of workflow management systems, we propose a novel XML-based profile model and sophisticated filtering methods that take schema information in the XML documents into account. We describe several indexing mechanisms and search algorithms for performing efficient filtering of XML documents for large-scale information dissemination systems and examine their performance across a range of document, workload and scale scenarios.

Back to the Summer 2000 dbchat index.