Research On Energy-saving Virtual Machine Migration Algorithm For Cloud Data Center

The Cloud Computing Center provides users with a new convenient IT service model that responds to various needs dynamically and schedules computing resources rapidly.With the demand for cloud computing services increasing continuously,the scale of data center is growing.As a result,the data center is confronted with a severe problem of high energy consumption.Therefore,it is important to build a green data center.The live migration of VMs technology is widely used in energy management,and plays an important role in energy-saving management of large-scale data centers.Migrating virtual machine to consolidate services on low-load servers and reducing the number of active servers,is one of the effective method to reduce energy consumption of data center.Based on that,take the characteristics of fat-tree network topology into consideration,we consolidate network traffic to cut down the number of active links without exceeding the link bandwidth capacity.As the decrease of active links,it is possible to reduce the number of active switches.The thesis focuses on reducing server power consumption,switch power consumption,and network communication overhead by migrating virtual machines.The research of this thesis includes:1.Propose the expand quantity gap energy saving virtual machine migration algorithm(EQG).the algorithm migrates virtual machines in batches according to the network traffic,which is light or heavy.First,migrate virtual machines with heavy communication traffic to close servers,which means the servers have shortest route distance.According to priority,it is best to migrate them to the same server,and then the severs directly connected to the same edge layer switch,and last the same pod.After that,the virtual machines will be migrated in descending order of their mutual communication traffic.The result of the migration is to expand the quantity gap of virtual machines between different server sets.The simulation results show that the algorithm has obvious effect on energy saving.2.The EQG algorithm is not so significant for the traffic aggregation at the edge layer as it at other layers.Therefore,a multi-level K-means virtual machine grouping algorithm(MLKG)is proposed to solve this problem in the fourth chapter.According to the clustering algorithm,the MLKG algorithm adopts an improved K-means algorithm to group the virtual machines three times.The first cluster aims to maximize the internal communication traffic of the server and reduces the communication traffic between the servers;the second cluster aims to maximize the communication traffic inside the server subset directly connected to the same edge layer switch;The third cluster aims to maximize communication traffic within the same Pod,reducing communication traffic between Pods.The experimental results show that after the MLKG algorithm is run in the data center,not only the energy consumption of server and switch are reduced,but also the total communication load of link in every layer is greatly reduced.