A Service Network Construction Method For Mass Customized Requirements

With the rapid development of the Internet and cloud computing technologies,software services available on the Internet have been increasingly diverse. The scaleof customers whom service providers are faced with grows rapidly, and thepersonalization degree of customer requirements on services becomes higher, too. Inorder to meet the customized requirements, service providers need to constructfully-personalized composition solutions; on the contrary, to reduce the cost ofservice delivery, it requires that a standard solution should be used to meet themultiple requirements of customers. There is a compromise between servicepersonalization and the cost of services.In the field of service computing, there have been many studies on this problem.However, traditional service composition method is mainly single-requirement-oriented. Even though it can adapt well to customization, the cost is extremely highwhen faced with massive customers. Software as a Service (SaaS) uses variablepoints configuration mechanism by leveraging one software service instance to meetmulti-tenant requirements, but the available service functions are mainly from SaaSprovider’s development, operation and maintenance, thereby the customizablecapability is limited. To resolve the conflict between customization and the cost ofservices, this thesis uses the service network to satisfy customized requirements andfocuses on the study of service network construction method:(1) Firstly, a formal definition of service network is given and a comparativeanalysis on service network and traditional service composition methods and SaaS ismade. The mathematical model of service network construction problem isestablished;(2) A service network construction method for mass QoS-oriented customizedrequirements is presented. Massive requirements are sorted by the potential benefits,and the requirements with high potential benefits are first considered. With the coverrelationship between service processes and requirements, heuristic rules are used toconsider the satisfiability of follow-up requirements. The goal is to satisfy morerequirements with less service processes and reuse the services which have occurredbefore as far as possible, so as to satisfy customer’s QoS personalized requirementand at the same time, to achieve cost-effectiveness. The effectivess of the proposedalgorithm is validated by a set of experiments.(3) A service network construction method for mass function-orientedcustomized requirements is proposed. Considering the similarities between massivedistinct requirements, progressive superposition strategy is employed to continuously enhance the solution that has been already constructed. The goal is touse the least candidate services to meet as the most requirements to achievecost-effectiveness. The method also makes use of grouping strategy to reduce thealgorithm time complexity. The advantages that the proposed algorithm havecompared with other strategies are validated by experiments.(4) A service network construction method for mass QoS-and-function-orientedcustomized requirements is presented. It makes use of the two algorithmsconclusions mentioned above to construct the service network, i.e., first building aservice network prototype, then useing the method of sorting key nodes to optimizethe service network nodes in order. The goal is to maintain high service reusabilityand avoid a waste of the use cost of services at the same time, so as to reduce theconstruction costs and the use cost. The algorithm is also validated by experiment.