The Research On Virtual Machine Scheduling In Cloud Computing

The concept of cloud computing comes from the concept of gird computing, distributedcomputing and parallel computing. It changes the traditional serve mode and provides a newbusiness mode by combining with the Internet. Currently the resources of the cloud computingare the virtual machines. Cloud computing virtualizes different hardware resources in the datacenter by virtualization technology. When the hardware resources forms the virtual resource pool,the virtual machines are deployed for being used transparently. The users only know that thetasks are carried out in the cloud, but they don’t know exactly which the tasks are carried out at.Job scheduling and resource scheduling are two key technologies in the cloud computing. Ascloud computing’s commercial characteristic makes it focus on user’s Quality of Service,virtualization technology of the cloud computing makes job scheduling and resource schedulingsignificantly different from the parallel and the distributed computing.With the expansion of the scale of data center and the increase of number of users, makingthe virtual machines in the cloud be used efficiently and making tasks not wait for long time forbeing finished rapidly become the important issues of cloud resource scheduling. Researchingthe scheduling of the virtual machines is significant for researching cloud resources management.The issues of scheduling of virtual machines in the cloud mainly researches how to allocate thevirtual machine to user’s tasks for meeting the user’s request of QoS, which improves cloudresources utilization rate, reduces computing time and assures the Service Level Agreement.This paper has done a research in the scheduling of virtual machines in the cloud computing.Based on previous researches, major research works and innovative points in this paper are:1. This paper analyzes the current status of cloud computing and the scheduling of virtualmachines in the cloud computing. It also systematically analyzes the conception, classification,application scenarios, major technologies and Map/Reduce distributed frame of cloud computing.With the careful analysis of the characteristic of cloud resources, author summarizes the model,the goal, characteristic, traditional scheduling algorithms of virtual machine scheduling. It alsointroduces a simulator called CloudSim, which is applied to the resource scheduling of clouddata center.2. This paper analyses the Map/Reduce frame carefully. Based on the Map/Reduce frame,author proposes virtual machine scheduling algorithm based on ant colony optimization (virtualmachine scheduling algorithm based on improved ant system and virtual machine schedulingalgorithm based on improved ant algorithm). In most current ant algorithms ants don’t communicate with each other directly, only rely on the pheromone exchange. The proposedvirtual machine scheduling algorithm based on improved ant algorithm brings in ant-meetmechanism, which can be used for finding the virtual machines in the cloud quickly through thedirect communication between the ants. Simulation results demonstrate the proposed virtualmachine scheduling algorithm based on improved ant algorithm can find the virtual machinesfaster than the proposed virtual machine scheduling algorithm based on improved ant system andassures Service Level Agreement.3. This paper analyses the traditional trust-driven TD Max-Min algorithm, trust-driven TDMin-Min algorithm and energy-aware virtual machine scheduling algorithm. At present existingscheduling algorithms of trust-driven ignore energy requirements. An energy-aware andtrust-driven virtual machine scheduling algorithm is proposed. The proposed algorithm meets therequests of energy-aware in data centers while ensures the user task performance. The proposedalgorithm maps virtual machine and tasks by trust-driven mechanism between task and virtualmachine, deploys the virtual machine by initialization algorithm of virtual machine and migratesthe virtual machines by minimize migration algorithm. Simulation results demonstrate theproposed algorithm outperforms the trust-driven TD Max-Min and TD Min-Min algorithms interms of the total trust utility, average trust utility and the level of Service Level Agreementviolation while consumes less energy.