The Key Technology Research On Complex Event Processing Engine

With the appearing of event-driven architecture, widely using of automation of businessactivities, and a growing need of information monitoring systems, a lot about the legal,contractual and operable rules are added to the distributed systems, which leads to anincreasing generation of events. This also increases the demand of systems to manage andprocess event automatically. Complex event processing technology caters to this demandperfectly. Whenever events occur, complex event processing engine automatically capturesthese low-level events for high-level knowledge to identify all the events in the systems.Complex event processing engine has two core technologies, one is the rule definitionand the other is event detection. The difficulty of Rule-making lies in identifying theparameters and adapting to the changing needs. Current solutions rely on the domain expertsto adjust the rules, which make the rule-making become a major obstacle for complex eventprocessing applications. The key of event detection is to accurately and quickly detect theevents. In order to achieve this objective, many detection models and their extended modelsare applied in complex event detection. However, these models haven’t considered the timeconstraints of the events fully. They only consider the time constraints of the instantaneousevents other than that of interval events, which leads to the limitations of complex eventsapplications.For the shortcomings of the research on the two core technologies, the paper presents thesolutions. At first, we propose an adaptive rule model which was built based on the idea ofkalman filter. This model predicts parameters by learning the given events which haveoccurred, and adjusts parameters according to the indirect feedback from the domain experts.This paper proposes the implemental algorithm, and designs two groups of experiments whichstudy the learning capability of the model as well as the number of intervals’ influence on themodel, the results prove the feasibility and effectiveness of the model. Then the thesispresents a complex event detection model based on interval time constraints. The modeladopts the automation combining with the instance stack to identify the event sequence.Based on the model, we propose a detection algorithm. In order to further improve theefficiency, the paper proposes the idea of pushing the constraint down and adopting thepartition instance stack, and gives the optimization algorithm. Finally we design and make three groups of experiments, both from memory usage and execution time aspects to verifythat the algorithms are feasible and the optimized algorithm performs higher efficiency.