Parallel Content Matching In Publish/Subscribe Systems

Suhas Doijad, Medha Shah

Abstract: Publish/Subscribe (Pub/Sub) system is the more optimized implementation of complex event processing system, in which user gives the interest (subscriptions) and some other party publishes the event (e. g. Stock quotes). The main functionality of Publish/Subscribe system is to send these events to subscribers whose subscriptions are related to the events. So the core of this system is to match the events with the subscriptions. As two phase algorithm is good in spatial locality and performs better, its implementation is proposed in this paper. The process of matching is parallelized using current cheaper and commonly available multi-core systems. Also, different parallelization strategies are used to improve the throughput and to reduce the matching time. Performance of the system is measured by taking processing time as a parameter. The results show that, the proposed technique gives more throughputs along with speedup of 3.4x and efficiency of 40 %.

Keywords: Publish, Subscribe system, Multi-core processors, Matching algorithm, Event processing, Stock quotes