Subscription summarization: a new paradigm for efficient publish/subscribe systems

We contribute a new paradigm for publish/subscribe systems. It is centered on the novel notion of subscription summarization. We first present the summarization structures for a broker´s subscriptions and accompanying algorithms, which operate on the summary structures to match incoming events to the brokers with relevant subscriptions and for the maintenance of subscriptions in the face of updates. Second, we present novel algorithms for efficiently propagating subscription summaries to brokers. Finally, we present a novel algorithm for the efficient distributed processing of incoming events, utilizing the propagated subscription summaries to route the events to brokers with matched subscriptions. We study the performance of our contributions, comparing them against a baseline approach and against corresponding techniques employed in a well-known event-based distributed system. Our results show the significant performance gains introduced for both the subscription propagation and distributed event processing tasks.