Font Size: a A A

Research On Multi-tenant SaaS Application Optimizing Deploy Algorithm

Posted on:2014-08-21Degree:MasterType:Thesis
Country:ChinaCandidate:Z F CaoFull Text:PDF
GTID:2268330422951512Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
Software as a Servic(eSaaS)is cloud computing architecture at the applicationlayer. Under this mode, customers do not need to purchase a complete softwaresystem, and not need to be equipped with corresponding hardware and maintenancepersonnel, only need to rent application software through the Internet on-demand. Inmulti-tenant SaaS model, multiple tenants share an application instance. As SaaSproviders can simultaneously reduce operating costs and customer softwarepurchase cost, it’s been gradually adopted by major software providers. How toplace tenants and deploy applications to make a reasonable system resourceallocation and utilization, provide services for more customers with the leastresources to obtain greater profits is a major problem of SaaS providers. In order tosolve the multi-tenant SaaS optimization deployment and resource allocationproblems, this paper studies the following aspects:First, aim to issues of static optimization deployment, a resource calculationmodel based on the criteria response time is presented to calculate the resourceconsumption of a user and SaaS applications on the server node. Description ofstatic optimizing deployment issues is proposed in both cases of SaaS provider rentinfrastructure and use its own infrastructure. For different problems we provideindividual model, and then use a genetic algorithm based on greedy strategy andgroup coded genetic algorithm to solve the static optimizing deployment strategyselection problem in two cases and verify them through simulation experiments.Second, aim to dynamic optimization deployment issues, a detailed descriptionof the problem and the problem model are proposed. Then we analyze the differenceof Euclidean distance and KLD (Kullback-Leibler Distance) distance in thecalculation of the distance between the requested resource vector and server surplusresources vector. We design two dynamic heuristic algorithms according to thedistance calculation method, the basic minimum remaining resources algorithmusing Euclidean distance and minimum remaining resources algorithm based onKLD distance, and both are verified and compared through experiment.Then, aim to resource consumption model and its adjustment problem, wedesign a monitoring model for acquiring runtime resource consumption of serverand SaaS application, then sends back to the resource monitoring center for analysisand processing for correction of resources consumption model of SaaS applications.Finally, aim to resource monitoring and tenant optimizing deployment of SaaSapplication issues, we design and implement a prototype system from thearchitecture, functional modules, databases, and other aspects.The system is equipped with functioning of tenant registration, deployment strategy generation,resource monitoring and other functions. It’s a simulation system of deployingbusiness process, with the integration of basic information management, monitoringcomponent, the algorithm functions and e.t.
Keywords/Search Tags:SaaS application, tenant placement, resource model, static optimizingdeployment, dynamic optimizing deployment
PDF Full Text Request
Related items