Deadline-Aware Event Scheduling for Complex Event Processing Systems

Abstract

Complex event processing (CEP) plays an important role in developing responsive stream processing applications, with emphasis on “Velocity” from Big Data perspective. However, in emerging applications with heavyweight query requirements, the big rule set and the event coupling relationship could result in a complicated event graph. Meanwhile, state-of-the-art graph-based event processing approach employs the depth-first strategy and assumes that all complex events are equally important, which are more likely to reduce the scalability. In this paper, we focus on assigning priorities to complex events heuristically for the event graph model. The proposed priority dispatching policies take the following aspects into account: the topological particularity, the relative deadline and the dynamic process of event correlation, with the goal of minimizing the average response time. Furthermore, a priority-driven event scheduling strategy is presented, which aims to support real-time reasoning requirement. Finally, the experimental comparison with the standard graph-based event processing technique shows that the proposed event scheduling can yield a substantial improvement in the specific performances.

This work is supported by National Natural Science Foundation (NNSF) of China under Grant 61174190, 61070048.