The Target-driven Automatic Service Composition Algorithm

In service-oriented computing,service providers encapsulate the heterogeneous resources into atomic services and provide them to users.In order to solve the contradiction between the complex needs of users and the limited functions of a single atomic service,service composition(SC)technology combines multiple atomic services into new services(i.e.composite service)realizing the reuse and appreciation of the service.To generate composite service that meets user need and has better quality of service(QoS)from multiple atomic services with similar functions,service composition technology based on QoS perception(QSC)is developed.Generally speaking,the QSC technology can be classified into two categories.The static QSC technology generate composite service based on predefined abstract service chain by selecting corresponding atomic service,while the dynamic QSC technology can generate service chain dynamically and select atomic service based on user requirement.With the development of pervasive computing and other related technologies,the category and quantity of atomic service are greatly increased.Meanwhile,the diversification of user demand leads to the increase of target types of QoS.At this point,the static QSC technology is limited by the predefined abstract service chain and cannot adapt to the increase of service type and diversified demand.The existing dynamic QSC technology is mostly based on the planning graph technology,which causes the high cost of service portfolio when services are too many,and the increase of QoS target type will also weaken its ability to optimize the QoS.To solve the problems mentioned above,this paper designs and implements a target-driven automatic service composition(TASC)algorithm.Firstly,a temporal target sequence model is proposed which transforms the process of solving QSC problem into the process of generating the related temporal target sequence.An intelligent planning algorithm is designed and realized based on the Yahsp2 planner to extract the composite service from the target sequence.The QSC problem was divided into several sub-problems by handling with the target sequence in a piecewise way,which reduces the complexity of the problem effectively.In order to optimize the QoS of composite service,a multi-objective genetic algorithm is designed and implemented,which takes the target sequence as the individual chromosome.It generates and optimizes the target sequence through population evolution.A hierarchical adaptive function is designed for multi-objective QoS realizing the QoS optimization of composite service.Finally,the test data from WSC2009 public data set shows that the TASC is better than to other existing methods in the case of a large number of services.