Content-based Publish/Subscribe (CPS) systems can e- ciently deliver messages to large numbers of subscribers with diverse interests and consequently, have often been considered an appropriate technology for large-scale, event-based applications. In fact, a signicant amount of existing research addresses the issue of providing scalable CPS services [3, 8, 7, 11]. In these approaches, scalability and high performance matching is often achieved by taking advantage of similarities between subscriptions. However, even though such optimization techniques are widely used, no model has been developed yet to capture them. Such an abstraction would allow CPS matching algorithms to be studied, analyzed, and optimized at a more fundamental and formal level. In this work-in-progress paper, we present the initial results of our work towards modelling and analyzing matching optimizations frequently used by CPS systems. Using our proposed model, we nd that probabilistically optimal CPS matching is possible in certain types of subscription sets and that there is also a non-obvious upper bound on the expected cost of some subscription sets. We also provide experimental results that support the model proposed and studied in this paper.
Readers who enjoyed the above work, may also like the following: