Research On The Virtual Machine Placement For Multi-VMs Interaction In Cloud Computing

Cloud computing reduces the expenditure of enterprises to buy big servers with its cheap hardware equipment services and high quality service experience,and enterprises do not need to spend any money to manage and maintain servers with the automation management system.More and more companies migrate their services to the C loud Computing.Enterprises can require resources on-demand through the Iaa S provided by C loud Computing.The cloud center will allocate enough resources according to the requirement.Iaa S allocates virtual machines which named cloud server to users.The controller searches for the optimal host which is called the target host for the virtual machine.This is the issue of virtual machines placement.It is not independent each other to the business owned by an enterprise.O ne service will give some support to other services.Those virtual machines hosted the services build a virtual Lan.It is the issue that we explore to choose the destination hosts for those VMs which interact with others.This problem is a Multidimensional Packing Problems.Also it is a resource optimization problem.The aim of this paper is to minimize the power consumption and the loss of resources.At the same time,minimize the cost of communication is considered because of the virtual local area network composed by VMs.These targets can be as the targets of a multi-objective optimization problem.Recent studies sho w that the heuristic swarm intelligence algorithm has great advantages in solving multi-objective optimization problems.In order to implement the deployment of Multi-VMs interaction rapidly and efficiently,and at the same time,take into account of the influence of the network overhead,resources wastage and power consumption,it take some measures to realize the target in this paper.First,group the VMs requests reasonably according to the relation between VMs.After this,then resource requests reassembled into multiple virtual machines sets.Then this paper use particle swarm optimization algorithm,which combines a mutation operator.With the operator,HGM-PSO presented in this paper can get the global optimum solution,and avoid local optimum.After grouped,the influence caused by any other request uncorrelated will be eased.The resources allocated by the monitor will more reasonable and scientific.According the fitness function,the algorithm output a solution that has the maximum fitness value.I n this paper,we study through this approaches.As suggested above,in order to implement the algorithm we proposed,this paper utilizes C loud Sim as the simulation platform for cloud computing,and FAT-Tree topology as the bottom architecture of the data center.The algorithm not only can avoid the case of local optimum,but also can prevent the effect coming from different VMs set.Using SDN as the network architecture,and examine the delay between virtual machines in different strategy.First of all,in order to program the algorithm proposed,study the simulator Cloud Sim,and master its basic architecture and functions of the platform.Subsequently,research Swarm Intelligence optimization algorithm,especially Particle Swarm Optimization and improved.Then,learn the principle and the controller of SDN,and the FAT-Tree topology.Finally,realize PSO in the platform of C loud Sim,and test the network delay with virtual network environment.Based on the above task,the Heuristic Grouping Mutation Particle Swarm Optimization algorithm is proposed in this paper.According to the available resources in the C loud Computing,it gives the optimal approach which can ensure minimum communication overhead,maximize resource utilization for the VMs set.And then,ver ify the different strategies about network delay in virtual network environment.According to the test results,this algorithm can guarantee the minimum network delay and a global optimal solution through grouping and mutation operator.