Research On Correlation-aware Manufacturing Cloud Service Matching & Composition Approach

Cloud manufacturing transforms scattered resources and capabilities supplied by entity enterprise into integrated manufacturing services, which can be realized through composing different services. And cloud service matching and composition are two research focuses in the field of cloud manufacturing research. As we all know, the service matching is a prerequisite for service composition, and they are both indispensable for providing users with satisfactory service. However, there is a question in current study: If the correlation is not considered in the matching process, services that have both poor QoS(Quality of Service) and correlation will be filtered out during composition process, As a result, the QoS of composite service is not as high as expected. Similarly, considering correlation only in composition process, the composite solution is also unsatisfactory, because services that have correlations have been left out during matching process. The problem can be described as how to consider the influence of correlations both in matching and composition process. To solve the above problem, the main works of this paper are summarized as follows:First of all, we need to extend the description model of manufacturing cloud service. On the basis of previous five-tuple description model, we add correlation description including quality correlation and statistic correlation, and use the six-tuple as extended service description model.Next we design two service reservation algorithms on the basis of the extended model: one is QoS-based service reservation algorithm which realizes the reservation of high-performance services, and the output of this algorithm is traditional candidate service set. Another is correlation-aware service reservation algorithm, which realizes the reservation of services that have both low QoS and correlation. And new candidate service set consists of outputs of these two algorithms.Then in the Matlab environment, through goal programming model and genetic algorithm, we do the optimization solution respectively on the traditional candidate service set and new candidate service set.At last, the simulation results directly verify rationality and effectiveness of extended model and reservation algorithm, and indirectly answer the question that if it necessary to consider the influence of matching and composition simultaneously raised at the beginning of this paper. In addition, the simulation also show the conclusion that the higher the ratio of services that have correlation to total services is, the more significant the proposed model and algorithm are.