Parallel Processing For Mobility-aware Complex Event

Complex Event Processing(CEP)is a technology that support real-time analytics of the events by using a set of event processing units called operators.In mobile envi-ronments,CEP system needs to update user's query range every few seconds base on his location.This kind of systems is called Mobility-aware CEP(MCEP)system.State-of-the-art MCEP systems mainly focus on distinguishing events that gener-ate from different query ranges.None of the works considers using parallel architecture to achieve higher throughput.In this paper,we introduce parallelism into MCEP and solve three additional challenges.First,mobility-awareness must be guaranteed under parallel architecture.Second,scalability is required to dynamically change the re-sources allocated to the system.Finally,load balance is required to balance the amount of each replica's input.In general,this paper makes the following three contributions:1.We propose a novel parallel architecture of MCEP.We use replica instead of operator as the event processing unit under our architecture.Replica partitions the output event stream it generates,and merges the events sent from preceding replicas into one input stream.We also propose a synchronize mechanism under parallel architecture to separate events from different query ranges.2.We propose a scaling strategy to perform transparent auto-scaling of parallelized MCEP systems.Also,we propose a dynamic load balance algorithm to balance the workload of different operator replicas during runtime.Auto-scaling and load balance use the snapshot of workload after every query range to detect operators that need adjusting.3.We conduct a series of experiments on virtual machines to evaluate the effective-ness and efficiency of our approaches.