Research Of Saas Resource Management Method Of Supporting The Intermittent User Behaviors

With the development of Internet technology, SaaS has arisen as a kind of innovative mode of software applications. Because of SaaS’s different application fields and ways of realization, user behaviors, supported by SaaS applications, are usually certain intermittent. This intermittency makes peaks or troughs of all SaaS services not synchronous, which will lead shift of resource consumption, uncertainty among different SaaS services and instability of QoS in the process of user behaviors. However, neither mature or cluster configuration methods nor accurate mathematic models was existed in the application field of SaaS resource management supporting the intermittent user behaviors. Therefore, how to solve the question has been an urgent and focused problem in scholars.In order to achieve efficient SaaS resource management of supporting the intermittent user behaviors, a SaaS resource management method of supporting the intermittent user behaviors was proposed in this thesis. Resource was deployed real-timely, dynamically, and rationally according to the number of concurrent requests based on the intermittent user behaviors to achieve the purpose of service performance and energy saving. Firstly, the SaaS resource management mechanism of supporting the intermittent user behaviors was proposed, the intermittent user behaviors’characteristics was analyzed, the SaaS resource management process of supporting the intermittent user behaviors was given, some important links like the services peak prediction, virtual machine performance analysis and start/stop control based virtual machine etc. Secondly, the resource usage of SaaS supporting the intermittent user behaviors was forecasted which was consist of two parts:one was to predict the concurrent request number of the single intermittent user behavior based on the Improved Grey Markov model, the other is to fit the concurrent requests and resource based on the RBF neural network, predict resources according to the prediction of the concurrent requests number and verify the effectiveness of the proposed algorithm by experiments. Finally, according to the number of concurrent requests, a SaaS resource management method based on the isomorphism virtual machine was adopted to manage SaaS recourse supporting the intermittent user behaviors efficiently, then verify the effectiveness of the recourse management by experiments.Compared with the existing resource management methods, this thesis presents the SaaS resource management mechanism supported the intermittent user behaviors in this thesis can adjust the virtual machine size according to the number of concurrent requests and resources usage to achieve a normal case in peak of resources and saving in trough of resources.