The Research On Complex Event Processing Method For Multiple Event Sources

With the rapid development of Internet of Things(IOT),enormous events data are produced every day.Complex event processing(CEP),which can be used to extract and identify meaningful information from the continuous data stream,becomes the key part of the IOT middleware.As event stream of IOT applications has many characters,such as massive,heterogeneous,distributed,and uncertainty,current event processing method can't handle such data efficiently.In this paper a high efficient distributed CEP architecture and method is proposed for multiple sources.Meanwhile,for the uncertain data processing problem,a complex event processing method over distributed probabilistic event streams is given.We further design a parallel algorithm to improve the performance 'of the system.The main works of this paper include:Firstly,there was a throughput bottleneck for centralized complex event processing system.In this paper we present a high performance distributed complex event processing system,which based on a distributed computing platform Storm.A parallel framework by operator for complex event processing system was proposed,the event stream parallel processing strategy on operator was given.In addition,we make some optimizations for the event processing algorithm,presents a distributed processing algorithm based on priority.Our experiment result shows these methods would achieve higher event detection throughput and retaining low detection latency.Secondly,In order to solve the uncertain event processing problem,a high performance complex event processing method over distributed probabilistic event streams is proposed.This method uses probabilistic nondeterministic finite automaton and active instance stacks to process complex event in both single and distributed probabilistic event streams.A optimized parallel algorithm is designed to improve the performance.Meanwhile,a query plan based method is used to process hierarchical complex event from distributed event streams.The experimental study shows this method is efficient to process complex events over distributed probabilistic event streams and have better performance and scalability than other methods.