Parallel Complex Event Detection Based On Regular Tree Pattern Matching

With the rapid development of computer technology,vast amounts of streaming data is continuously generated,transmitted and processed by network applications.Streaming data has the characteristics of huge amount,various types,fast flow rate,but low monomer value density.Therefore,looking for ways to get high-value density information from streaming data is a key to the development of big data era.Complex event processing is a technique based on the flow of events.It abstracts data into different types of events,analyzes the relationships between events through operations such as filtration,association,aggregation,and finally obtains the high-level events from simple events.Since a considerable amount of streaming data can be abstracted as the flow of events,the detection processing of streaming data can be abstracted as the detection processing of complex events.Thus,developing the complex event processing method with stronger event description capability and higher event detection efficiency has become one of the main ways to address the big data era's need.However,strong event description capability and high event detection efficiency is a contradiction for complex event processing.The reason is that multiple simple events which constitute complex event are often possessing both timing constraint relationships and structural constraint relationships,so high-level event pattern should provide rich constraint description capabilities.At the same time,the complexity of the detection algorithm is inevitably improved.In addition,the fast arrival of streaming data increases the overhead of computing and storage resources.This study proposes a complex event detection method "CEDPRT" by using the multi-core computing capability of computer,which provide a new way to solve the above problems.It uses "regular tree pattern" with strong description capability as the event pattern,so that CEDPRT can not only process users' detection requests with structural constraints,but also process the requests with timing constraints,so that the method has "strong event description capability".Through "automata" and "mapping",a complete regular tree pattern matching method by data-parallelism is proposed,which makes the computer's multi-core advantage fully utilized in the processing of streaming data.Through the step of "tree ordering",the matching task is decomposed,and taskparallelism is combined with data-parallelism to process regular tree pattern matching,so that the method has "high detection efficiency".Through experiments,the description capability and detection efficiency of CEDPRT are proved,and users can get deeper information more efficiently by using CEDPRT in network applications of big data era.