Towards prioritized event matching in a content-based publish/subscribe system

Abstract

QoS support is important for a large-scale content-based publish/subscribe (pub/sub) system to provide guaranteed service for clients with high QoS requirements. So far, great efforts have been dedicated to integrating QoS support into pub/sub systems. However, most work focus on providing QoS support on routing, without touching QoS support in event matching. In this paper, we propose the idea of prioritized event matching, aiming to integrate QoS support into event matching. We first point out the lack of time metrics that reveal performance detail of matching algorithms, leading to the definition of new time metrics. Through a series of experiments conducted in terms of new metrics, we discover the foundation for prioritized event matching. Finally, we realize prioritized event matching, called Pri-Rein, based on an existing matching algorithm and provide three design guidelines learned from the lessons in Pri-Rein. Extensive experiments are conducted to verify the effectiveness and efficiency of Pri-Rein and results show that Pri-Rein well achieves our design goal. We argue that the idea proposed in this paper can be generalized to matching algorithms that are used in cloud computing or complex event processing.