Performance Analysis Of Multilevel Indices For Service Repositories

The emergence of Service-Oriented Computing(SOC)provides a flexible solution for enterprise software integrations and distributed systems.While ServiceOriented Architecture(SOA)is an infrastructure of SOC,and Web service became the element of SOA.Web service has grown drastically in numbers during the recent years.Furthermore,the rapid development of cloud computing also promoted the development of service,a large number of services converges to a cloud platform centered service repository.Service discovery and composition are the two important research direction in the field of computing.Therefore,how to manage a large-scale services repository efficiently to promote the service discovery and composition is a very important and challenging problem.This dissertation first introduces the research background and significance of the service,research and introduced the research status of the service discovery and service composition,and on this basis,service index structure,inverted index and three kinds of deployment model of multilevel index model are studied in detail.The concrete research contents include:(1)Inverted index is widely used to store services in large-scale service repositories since it can reduce search space of service search.However,there are still many redundancies residing in the inverted index,which reduces the efficiency of service discovery and composition.By addressing the information redundancy problem,the storage structure of multilevel index model and three types of deployment models eliminated information redundancy in the greatest degree,and can effectively promote the efficiency of service discovery and composition.But for a specific service repository,Multi-level index model does not give a model which can be used to select the most efficient index of effective standards.To addressing this issue,this dissertation,first analyzes the performance of different index models structure,and service expectation function is given according to the characteristics of the repository and service,as the basis of service library to choose efficient service index model.And then a detailed formula derivation process of expected function of the five model indexes is given.This study then analyses the characteristics of each index model by optimizing the deployment plan,and the characteristics of the applicable service repository,etc.Finally,accuracy and reliability of the proposed optimization deployment scheme and expect function are validated by experiment.(2)In view of the key values optimization selection algorithm of multilevel index model carried on the preliminary discussion.First of all,this study carried on the thorough analysis and introduction for the key values selection principle of multilevel index model,and help understand through a variety of simple example.And then the study analyzes the problems existing in the key value selection problems.Finally,the retrieval performance of key value index structure optimization and the optimization condition has been evaluated.The results show the key value in the case of random and without any effects on retrieval performance of the index structure,prove that the problem existing in key optimal selection,and also verify the correctness of our preliminary investigation of the key value optimization selection algorithm.