Research On Mechanisms For Composite Service Reliability Enhancement

A composite service is usually composed of multiple component services belonging to different providers,and runs in an open,dynamic and varying network.Thus,the execution environment of the composite service shows greater uncertainty,leading to the difficulty to guarantee the reliability of composite services provision.Therefore,how to provide a highly reliable composite service becomes important in the field of service computing.In this dissertation,we study it and design desired method from three aspects:transactional composite services provision in network,non-transactional composite services provision in network,and composite services provision in the edge network.The following results briefly summarize the contribution of this dissertation:1)We propose a Service Level Agreement(SLA)aware fault-tolerance approach,so as to provide highly reliable transactional composite service in the network.Firstly,it employs the Finite State Machine to monitor the execution status of the composite service,and it can rollback to the state that can be compensated when a fault occurs.Then,it adopts the Monitoring Automata to monitor the attributes of SLA.Finally,when a fault or SLA violation occurs,an improved differential evolution algorithm is triggered to quickly calculate the optimal recovery plan,which ensures the reliability of composite service provision is highly.The experimental results with the real-world dataset show that the proposed approach is superior to other approaches in terms of the service repair time and quality of composite service.2)We design a fault-tolerance approach with service replication to provide highly reliable non-transactional composite service in the network.Firstly,the QoS with different dimensions is normalized by using the linear transformation.Second,the best-worst method is adopted to calculate the user's preference weights for different QoS attributes.Finally,a multi-attribute decision-making method is employed to comprehensively evaluate all of the candidate component services,and a component service with a high score is selected to replace the faulted one.Numerical results show that the proposed approach achieves higher quality of composite service and shorter service repair time.3)We propose a lo c ation-aware composite service provision approach in the edge network.Firstly,it decomposes the complex dynamic optimization problem into several static optimization problems.Each static optimization problem solves the problem of a certain location within a certain time limit,that is,the optimized intermediate solution is obtained.Moreover,the multi-population distribution estimation algorithm is used to implement the static optimization process,so that the obtained intermediate solution is closer to the optimal solution.Then,the optimal solution can be obtained quickly by reusing the intermediate solution.Finally,the roulette mechanism is used to increase the reuse probability of the optimized intermediate solution with high fitness.The experimental results based on the real-wor-ld dataset show that the proposed approach achieves a higher reliability,a better stability and a shorter execution time compared to other solutions.