The Research On Uncertain Event Complex Event Processing Method For Internet Of Things

In the 21st century,as technologies of perceptual recognition develops,devices of information generation begin to accurately sense,measure and monitor the physical world in real time.On account of networking requirements of the physical world and expansion demands of the information world,the Internet of Things as a new type of network emerges.A main problem to be solved of wide apply of The Internet of Things is the information processing part.Moreover complex event processing just is one of the core task of Internet of Things middleware.Complex event processing based on IOT environment faces many challenges,One is how to efficiently detect the real-time event flows of large number and high speed.The other is how to accurately compute the uncertain results caused by uncertain event source.In order to solve above challenges of complex event processing,this paper presents a method USCEP to detect uncertain event streams,and then we expand it to improve performance in field of distributed and parallel precessing.Details include:(1)USCEP is uncertain complex event stream processing model,and we defines its own query language and operators.This model assigns buffer list for each node of matched tree to buffer the historical matching records,which makes different matching process can share the intermediate results,and greatly increases the number of the output complex event.When each operator node generate new records,they will use mechanism of conditional probability chart/table to calculate their new probability.In order to improve performance in multiple registration queries processing,we use relational query table to reduce the useless query.The experiments show that USCEP has better performance than the contrast method when it processes uncertain event stream.(2)The distributed and parallel version of USCEP is built on the distributed real-time stream processing platform Storm.Faced with features of distributed applications,we describe our entire system architecture in detail,and the research priorities is how to make global CEP and local CEP efficient parallel.So,the methods that pipelined hierarchical parallel and local grouping parallel are proposed.Pipelined hierarchical parallel is that all middle operator node in matched tree are separated into single logical operation node,and then different logical operation connected each other by structure of tree to reach the form that pipeline-type.The content of local grouping parallel is that multiple instances and input data average grouping.After input data is separated into multiple instances,all data will merge in boundary bolt so that processing in parallel and independently.Finally,experimental results show that the method can efficiently handle massive uncertain event streams.