Study On Graph Matching Based Tenancy Recommendation Method In PTC-SaaS

With the development of cloud computing, SaaS (Software as a Service) has been widely applied and researched as a business model of service delivery. An application based on the same infrustructure is usually termed multi-tenancy SaaS if it offers common services to multiple tenants simultaneously by the way of renting on demands. How to customize persernized tenacies according to user requirments has become one of the hotspots in the research field of multi-tenancy SaaS. Redeploying and building a new tenancy will increase the cost of SaaS development and mantainance during the course of customizing. For this reason, how to recommend tenancies that meet user requirments on the basis of history SaaS deployment expriences becomes a key issue in terms of reducing cost.To solve these problems, this thesis proposes a kind of graph matching based tenancy recommendation method in multi-tenancy SaaS, which recommends tenants with proper deployed tenancies to raise the tenancy reuse and reduce the cost of tenancy deployment. This thesis formulates a framework of personalized tenancy customization in multi-tenancy SaaS (PTC-SaaS), bases on which it elaborates the complete personalized tenancy customization process including tenancy demand customization, effectiveness test, tenancy recommendation, tenancy deployment as well as tenancy implementation, and then focuses on tenancy recommendation stage. To solve the large target solution space and low matching efficiency problem, a tenancy index table and a filter algorithm are presented. The algorithm can effectively reduce the solution space and dramatically improve the efficiency of tenancy recommendation. In addition, an error-correcting subgraph isomorphism based graph matching algorithm, together with the methods of measuring matching between business activities as well as transitions is proposed. The algorithm can search out the tenancy most similar to tenancy demand in the aspect of tenancy business from the recommendable tenancy set, which could achieve the tenancy recommendation. Finally, this thesis presents implementation details of the algorithms proposed in the tenancy recommendation stage, and analysis of factors which affect the quality of the error-correcting subgraph isomorphism based graph matching algorithm in experiments, and verifies the importance of the tenancy filtering algorithms on improving the tenancy recommendation performance.