Enhancing performance and expressibility of complex event processing using binary tree-based directed graph

Abstract

In various domains, applications are required to detect and react to complex situations accordingly. In response to the demand for matching receiving events to complex patterns, several event processing systems have been developed. However, there are just a few of them considered both performance and expressibility of event matching as focusing only on performance can cause negative effect on the expressibility or vice versa. This research develops a fast adaptive event matching system (FAEM), a new event matching system to improve expressibility and performance measures (throughput and end-to-end latency). This system is designed and developed based on a novel binary tree-based directed graph (BTDG) as a unified basis for event-matching. The proposed system transforms a user-defined query into a set of system objects including buffers, conditions on buffers, cursors, and join operators (non-kleene and kleene operators) and arranges these objects on a BTDG. Provided BTDG the enhancement in performance of non-kleene operators applied through developing a batch removal method to remove the events that are located out of time-window, and an actual time window (ATW) which can improve performance of event matching. To improve performance of kleene operators, this research introduces a twin algorithms for kleene operator which is match to BTDG. These two kleene algorithms apply grouping on events and reduce the number of intermediate results and apply combination algorithm in final stage. Transformation of queries containing join operators into BTDG enhances the expressibility of the proposed CEP system.