INFOrmation Dissemination (INFOD) middleware

ABSTRACT

The increasing volume, diversity, and complexity of resources and data continues to raise challenges in the sharing and dissemination of information. Although publish-subscribe systems have improved the ability of users to exchange information, we argue that these systems have been either restrictive or simplistic - relying significantly on narrowly defined channels for data exchange. The Information Dissemination (INFOD) approach we present here introduces a flexible and dynamic framework for brokering information in publish-subscribe systems. INFOD enables communities of interest to themselves constitute and use vocabularies for describing their interests as well as their capabilities (available information). Publishers, consumers and subscribers are real-world entities characterized in terms of vocabularies and their interests as constraints within an INFOD registry. Subscribers define subscriptions primarily as XQuery constraints expressing events of interest at candidate publishers. Subscriptions also specify constraints on the run-time data that must be disseminated to specific consumers. Entity descriptions, property constraints, and subscriptions comprise the metadata information that INFOD uses to associate and link entities within a community. We refer to the process of associating entities in INFOD as mutual filtering, which we realize with a three-way join across publishers, consumers, and subscriber entities. We demonstrate and evaluate the INFOD approach in an emergency response use case scenario that employs INFOD to support changing event dynamics and varying publisher-consumer-subscriber requirements. We use industry standard technologies and present system performance results for the mutual filtering steps for a variety of subscription constraints and broad classes of publisher, subscriber, consumer property descriptions.