Deadline-aware complex event processing models over distributed monitoring streams

Abstract

Complex event processing over data streams has raised a lot of attention in the database community, especially due to the development of ubiquitous data sensing equipments. Specially, some mission-critical applications have deterministic QoS requirements (e.g., event delay upper bound), in order for the results to be useful. However, most existing complex event processing methods only provide best-effort services with the statistics QoS, such as average response delay, which cannot guarantee more detected composite events that are responded under the deterministic QoS requirements. Therefore, it is quite imperative to provide a kind of deterministic QoS based processing model for these event-based applications. In this paper, we focus on the most important QoS metric in complex event processing monitoring applications, namely, deadline. First, each distributed monitoring stream is modeled separately as a stochastic process according to monitoring objects’ arrival situations. Then, the processing is modeled according to a generalized automata model. Moreover, a sophisticated time-cycle based framework is proposed based on the arrival and processing models to output more composite events under the deadline constraint. Finally, the experimental evaluation is made to validate the correctness and efficiency of our model for the deadline-aware complex event processing in the monitoring scenarios.