Service Selection And Negotiation For Service Composition With QoS Correlation

Service Oriented Computing is a new computing paradigm that utilizes services as fundational elements for low-cost and rapid application developing. As the dynamics and unpredicablity of service computing, Quality of Service (QoS) Guarantee in internet is a difficult and significant challenge. Currently, there are three categories of approaches for QoS guarantee for service oriented applications:software constructing approach, resource configuring approach and access controling approach. In this paper, we focus on the software constructing approach and research on the QoS guarantee for composite service.As the developing of the service oriented computing, there are more and more services emerging in internet. On the other hand, there are some correlations among the services, and the correlations have influence on the QoS of services. The existing approaches for QoS guarantee assumpt that QoS of a service is fixed and does not depend on other services. The service QoS description does not contain the information of service correlations and the dependency of QoS value on correlation, such that the actual QoS values deviate from the declared QoS values. And moreover, the deviated QoS values have a significant negative impact on the QoS-driven service selection and the QoS negotiation.In this paper, we address on the problem of the correlations among services and the dependency of QoS value on correlations. A novel QoS decription approach is proposed to model the correlations and the dependence of QoS values on the correlations. And then, we focus on the new challenge from correlations and research on the QoS-driven service selection and the QoS negotiation.Firstly, we propose a sematic-based approach for QoS correlation discovering. As there are correlations amongs services and the correlations have influence on QoS values, we propose a sematic-based approach to discover the correlation and the influence on QoS. A correlation-aware QoS model is presented to model the correlations and the influence on QoS values. However, a service provider can not descripe service according to this model, because this desription uses the information from other service providers. Therefore, an OWL-based service model is presented. This model only involes the service itself, such that a service provider can use it to descripe the services which are published by this provider.A correlation discovery approach is proposed to transform the OWL-based service description to the correlation-aware QoS description. We also use domain ontology knownlege to solve the probable problem of semantic heterogeneity in the OWL-balsed service descriptions, and a declared assertion to solve the problem that the correlations may change dynamically. The correlation discovery approach is evaluated by a real world example from Amazon (Amazon Infrastructure Web Service). This example illustrates the effectiveness of this approach.Secondly, we propose a QoS-driven service selection approach with correlation. As the accurate description of QoS is the foundation of QoS-driven service selection, the QoS of composite service can not be optimazed based on deviated QoS description. However, the existing selection approaches can not deal with the problem with service correlations. In this paper, we focus on the problem with service correlation. This approach models the QoS-driven service selection problem as 0-1 Integer Linear Programming, and then solves the problem through programming algorithm. However, the model becomes non-linear because of service correlations, such that service selection can not be completed in polynomial time. Therefore, we propose a theory to transform the model into a linear one. We evaluate the approach through the real world example from Amazon. The example illustrates the effectiveness of this approach. The QoS of composite service selected by this approach is better than the one selected by other approaches. Moreover, the performance overhead can be accepted through comparison.Finally, a histoty heuristic based QoS negotiation approach for composite service is proposed in this paper. When the QoS requirement is rigorous or the services which can be selected are scarce, the feasible solution can not be found directly. However, current QoS negotiation approaches can not get a maximal and fair solution for all parties. Moreover, these approaches are inefficient. In this paper, we model the QoS negotiation problem as an optimization model seeking maximization and fairness. Moreover, we improve the performance through a heuristic algorithm in which the feasible solution and the neighbourhood are generated through the heuristic historical negotiation data. Empirical study shows that, a maximal and fair QoS solution can be found through this approach, and experimental results indicate that the history heuristic based negotiation algorithm proposed in this paper is efficient comparing with the Simulated Annealing based negotiation algorithm.