Research On Optimization Methods For Virtual Machine Placement In Cloud Environment

Cloud computing is an important milestone in the development of distributedcomputing, as a commercial application, cloud computing has great developmentprospect. IaaS is an important service models in cloud computing，it could combinethe huge resource dispersed in different space into a resource pool, and thus makes themanagement and utilization of the resource more convenient. In IaaS，all the resourcesare provided in the form of virtual machines (VMs), and the optimal VM placementalgorithm has to be carried out in order to achieve high efficiency, small computingtime and low cost.In this paper, an algorithm for VM placement optimization is proposed under theresearch and analysis of the existing methods, and a corresponding model is also build.The main work and innovation of this paper are as follow:(1) The traditional methods for resource management are mostly about the innerVM resource optimization, however in our method the basic unit of resource is everysingle VM, and the model concerns about resource management of macro approach(2) The main object of this model is to achieve minimum cost and provide anreserve mechanism for resource access，when considering about the factor of cost andtime. This mechanism could offer a certain amount of resource and ensure some kindof utilization experience to the users. Also, this mechanism could improve the servicespeed, reduce the temporary resource transfer, and avoid network conflict. It couldreduce users’ cost and response time and promise high quality of service at the sametime.(3) In this model, the uncertainty of users’ demand is also taken into account, andthe uncertainty is then simulated by a stochastic variable with the probability ofnormal distribution to suit for the reality situation. In this paper, the conception of‘response time’ in VM placement is brought in. The response time is used as animportant factor in the process of optimization. The affection of response time to costis discussed according to the relationship between them. If the response time isrelated to cost, then the time affection is transferred into cost value. Otherwise, thetime affection is considered as a constrain condition. In the process of serviceproviding，cloud agent could choose different reservation mechanism for the usersaccording to different application situation. (4) In simulation，the model is solved by the method of two-stage stochasticprogramming. The results show that this model could efficiently simulate the resourceplacement problem in cloud computing and thus has great commercial potential.