Research On The Energy-Efficient Based Datacenter Resource Consolidation Mechanism

With the rapid development of the human society, environment issue increasinglyattracts peoples’ attention. The extreme weather which is caused by the global warminghas already warned and is warning human beings the importance of the greenenergy-efficiency. How to reduce carbon emissions has become one of the currentresearch focuses. Datacenter always has the huge demand for electricity since the earlybeginning of being founded. It’s very necessary to introduce the idea of greenenergy-efficiency into the field of datacenter. And with the escalation of virtualizationtechnology, cloud computing has been widely used. Making use of the "skin front-end"to access the "cloud" then enjoying the system function beyond the datacenter withoutany trouble of the hardware configuration has already made into reality. And thelive-migration technology also opens up a brand new approach to deal with the issues ofgreen energy-efficiency resource management beside the workload management.Firstly, author briefly introduces the historical background of cloud computing anddatacenter, and the related technologies trend, and raises the problem of the energyefficiency based datacenter consolidation mechanism.Secondly, because the consolidation mechanism needs the real-time monitoringinformation to support its judgment, author analyzes three common datacentermonitoring systems in the view of the actual deployment experience, and summarizestheir ability in physical machine power monitoring and virtual machine monitoringextension. At last the author provides two optimal virtual-machine monitor solutions atpresent.Thirdly, this thesis introduces the main sources of the datacenter energyconsumption in detail and two server energy consumption model. The author brieflydescribes the features and limitations of the existing hardware energy saving technology,and provides the guidelines of the energy-efficiency based on virtual-machineconsolidation mechanism. Then the author proposes the dual-thresholdenergy-efficiency mechanism，describes its logical structure and logical flow, especiallyhighlights the three key issues：threshold setting，virtual machine selection and the consolidation mechanism. Then author introduces three virtual machine selectionmechanisms.In addition, this thesis deeply studies the problem of virtual machine consolidation.In the fourth chapter author proposes three virtual machine migration algorithms fromdifferent perspectives including RP(Random Place), FFD(First Fit Decrease) andGCBFD(Group Check Best Fit Decrease) consolidation mechanism. The basic idea ofRP algorithm is a random placement policy. The key idea of FFD algorithm is totranslate the consolidation problem to the bin packing problem. The key idea of GCBFDalgorithm is to introduce the grouping-migration concept. After all, this chapter providesthe simulation of the algorithm. The result shows that the GCBFD algorithm has higherenergy-efficiency performance than other two algorithms.At last, the author studies the establishment of the above energy-efficiency systembased on the open-source cloud platform OpenStack, and proposes the modules functionand logical interfaces of the OpenStack based energy-efficiency cloud platform.