A Service Network Customization Method For Personalized Requirements

With the rapid growth of the IT technologies such as cloud computing and IOT,softwares and services are flooding the Internet. Customers’ requirements now can bemet by web services and their compositions. Meanwhile as the user groups grow larger,a wider variety of requirements needs to be met. And it’s become one of the researchareas in the field of service computing to meet the personalized requirements throughtcomposition of the services.At present there are mainly two methods on the issue of how to meet thepersonalized requirements：Mass customization and service composition. Traditionalmass customization methods such as SaaS, support flexible customization throughmodifying the configuration files. However this kind of customization system isessentially closed. They have limited service ability and scope. Traditional servicecomposition methods chiefly focus on customization for single requirement. They cansolve the customization issue for personalized requirements. But facing masspersonalized requierments, they are regarded as high-cost.To solve the above issue, this paper propose a customization method based on theService Network (SN for short). The SN tries to connect the heterogeneous andseparated web services in some protocol, forming an open, large network with strongcustomizing ability. When requirements come, the SN conduct the customizing methodto find the optimal solution (similar with the composition services in traditional servicecompostions) to meet the requirments. On the premise of an exiting SN, this paperconducts the research of the SN optimization issues in the following aspects:(1) Propose the concept model of SN and the corresponding mathematical model.Describe the presentation of personalized requirements and propose a practical case toshow how the customization is done with the SN.(2) Propose the method of SN customization for single requriment. First set up themathematical model, the optimization goal of which is to minimize the cost on thepremise of satisfying the customer’s requirements. Use ABC (Artifical Bee Colony)algorithm to solve the issue, and introduce the encoding method of food souce, theoptimizing process and the fitness calculation method in detail. Prove the effectivenessof the method through four experiments and analyze the several factors that affect SNcustomization performance. (3) Propose the method for SN customization with multiple concurrentrequirments. Put forward the concept of customization cost and use cost in the concernof the trait of concurrency. With the idea of reusing the chosen services, theoptimization goal is to lower the total cost in the premise that requirements are met. Weuse both the method MR-SNC which concerns reusing and SR-SNC which does not tosolve the problem for comparison. The results show that MR-SNC can effectivelyreduce the cost.(4) Propose the method for SN reliability enhancement based on insurancemechanism. The solutions produced by the above two methods do not contain anyredundancy. However in real life users often have more strict requirements forreliability so that redundancy is needed in the solutions to meet the requirements.Propose three ways (parallel redundancy, sequencial redundancy and path redundancy)to enhance the reliability of the solutions. In the conditions of having a customizationsolution and not having any, use heuristic method and ABC algorithm to solve theissue respectively. The essence of both methods is to find the optimal redundancescheme. The experiments prove the effectiveness of the algorithm.In the end, we design and realize a prototype system that supports the SNdeploying, customizing and executing. And introduce the architecture, system modulesand working process of the system.