Research On The Approach To Placement Of Tenants For Service Based Multi-Tenant SAAS Application

With the rapid growth of software technology, software application pattern is showing a trend of network, platform and service. New software architecture and relative technology has achieved rapid development with the support of distributed computing, parallel computing, grid computing and cloud computing. Against this background, as a new pattern of software service, SaaS (Software as a Service) is welcome by users and software vendors because of its advantage of on-demand using, no maintenance cost, easy to extend and so on.As a development mode, SOA (Service Oriented Architecture) guides business solution to set up, organize and reuse Web component. SOA could rapidly adapt the constantly changing business as it uses services as the basic units to develop applications based on the so-called "use, not own" pattern, and supports the rapid and low cost development in the distributed application.Multi-tenancy is one of the kernel technologies for the SaaS model. In the multi-tenancy SaaS applications based on SOA and cloud computing, tenants share the same software and hardware resources, and each tenant customize the software on-demand without impacting on each other by multi-tenancy. Multi-tenancy could effectively reduce the use cost to maximize the income with economy of scale.Service selection is hot research topic nowadays. A composite service consists of tasks and every task has a set of candidate services with the same function but different quality of service (QoS). A service selection means finding a Web service instance for each of the tasks in a composite service and Genetic Algorithm is the most widely method for service selection at present.One important objective of the multi-tenancy SaaS application is to maximize the income by sharing the software and hardware resources. In our multi-tenancy application, the number of servers is limited and so that the number of Web service instances deployed on servers is limited. In order to reduce the cost, one important problem is how to optimally place tenants to the limited number of servers to maximize the total supported number of tenants without violating their Service Level Agreement (SLA) requirements.To solve the problem above, we propose a hybrid approach to solve placement of tenants which uses a combination of service selection with genetic algorithm (GA), case-based reasoning (CBR), heuristic approach and resource consumption estimation model. The main idea is that TPS divides tenants into different categories based on the tenants’SLA and status of Web service resources in the service library, generates suitable number of execution plans by GA according to the SLA of different categories and the existing usable Web service resources, selects existing eligible execution plans by CBR, gets the most suitable execution plan by heuristic approach, places the corresponding tenant to this execution plan and finally recalculates the residual resources of services related.Under the manufacturing information technology platform, this paper presents and analyses the experimental results and develops a TPMS prototype system combined with supplier relationship management service (SRM). Summary of the constructed system and expectations of research on TPMS are described in the final part of the paper.