Customized information dissemination in distributed publish/subscribe systems

Delivering the right information to the right receivers in an efficient and scalable fashion has been an essential requirement in wide range of applications. Publish/subscribe (pub/sub) systems are an ideal infrastructure for providing such services to applications. A distributed pub/sub framework not only provides asynchronous and anonymous communication by decoupling publishers and subscribers, but also accommodates scalable information dissemination by distributing the content matching and routing load among several brokers.In this thesis we explore a distributed content-based pub/sub framework for customized information dissemination. We introduce a new distributed pub/sub architecture that achieves fast and reliable dissemination through a cluster-based broker overlay network. The proposed architecture provides fault-tolerance and load balancing in the distributed broker overlay network through subscription replication. We also investigate efficient subscription management and content matching techniques as these are primary operations that are performed inside brokers. Our work includes proposing novel content representation models and subscription management and publication matching algorithms based on the proposed model in a numerical content space where publications and subscriptions are defined over numerical values.Finally, we extend the traditional pub/sub infrastructure beyond the dissemination of simple content formats in order to efficiently accommodate customized dissemination of rich content formats such as multimedia content. This is done by intelligently placing content customization operators in the pub/sub broker overlay network to minimize the resulted communication and computation costs. Each of these contributions is accompanied by an extensive experimental evaluation that represents the effect of using the proposed contributions.