Research On Resource Management In Cloud Computing Datacenter

Cloud computing is a new commercial calculation service mode. By distributing thecomputational tasks on the resource pool which is consist of a large number of computers,users of cloud can get the memory space, computing power and information service fromanywhere and anytime according to their needs. Now, cloud computing has already beenwidely applied to the science computing, web search, biological information and energyindustry.With the rapidly development of cloud computing, the infrastructures of datacenter arebecoming increasingly large scale. Cloud datacenter can potentially house thousands ofservers, at the same time all kinds of applications and services run on them with differentresource requirements and dynamic workloads. At the same time, all these applications shareall the resources in datacenter. Virtualization makes sharing resource possible by enabling thecomputer system to migrate workloads between the different physical nodes and custom ofthe computing environments as application needed for adapting to changes. With theincreasingly large scale of the cloud computing, the growing management complexity and theinherent dynamism bring great challenges in resource management of cloud datacenter.Based on a systematic analysis of related works of datacenter resource management, thisdissertation focuses on virtual machine management in cloud datacenter. The major work is asfollows:1. Rack Based resource management heuristics for cloud computing datacenter. Firstly, wepresented the structure of the cloud computing datacenter, energy cost calculating model andmethods for datacenter performance evaluation. Then we proposed when to migrate virtualmachine, chose which virtual machine to be migrated and where to place virtual machine asthree key problems for resource management in cloud computing datacenter. To reduce theenergy cost and ensure the user service quality, we set the threshold for the server and rack.The system decides when to migrate by checking whether the workload of the rack or serveris out of threshold. For the problem of choosing virtual machine, system chose the minimumnumber of virtual machine to reduce the migrate cost. On the problem where to place VM, thesystem chose the node by the rack based heuristics. Experimental results show that theproposed cloud computing datacenter resource management heuristics can lower the energyand virtual machine migration cost while keep low violation rate of SLA.2. A multi-objective approach for resource management in cloud computing datacenter.The system considered three conditions for resource management which included ofdecreasing the power consumption, reducing the resource contention and improving thequality of service. The proposed managed approach focuses on the VM replacement problemdue to changes of application workload or the system conditions, for example, decidingwhether should migrate VM, choosing which VM to be migrated and where to place VM. Forthe problem of deciding whether should migrate VM, the system set low and high thresholdfor the different optimization targets, then used local regression techniques and sliding-window to decide whether should migrate VM. To choose the suitable VM to migrate, the system adopted different strategies to choose the VM according to the different optimizationtargets. On the problem of where to place VM, to balance the conflicts between differentoptimization objectives the proposed approach considered, the system used the TOPSISmethod to choose the host node. Experimental results show that the multi-objective cloudresource management approach based on TOPSIS can reduce the power consumption, getless violation of SLA and lower the resource contention while balance the conflicts betweendifferent targets.