Research On Muitple Virtual Machines Migration Scheduliog For Network Traffic Optimizarion In Cloud Data Cenrer

With the booming development of cloud computing, more and more enterprises and individual users deploy their business and applications in the cloud. Cloud data center, as a critical infrastructure, plays a mainstay role in cloud computing. With the development of new Internet online business and new technologies such as distributed computing applications, emerging services need a large amount of one-to-many and many-to-many communications between the servers. As a result, it raises tremendous east-west traffic problems. Virtual machine (VM) migration makes operation and maintenance more flexible, which is able to achieve server consolidation, fault tolerance, energy saving, QoS management, load balance and other optimization goals in cloud data center. However, VM migration traffic will increase network load even cause network congestion. Thus, cloud data center network traffic optimization has become an important issue to be solved.This thesis deeply studies the network traffic optimization in cloud data center from the perspective of multiple VMs migration scheduling under the new SDN architecture. By migrating virtual machines, we adjust the deployments of virtual machines running customer services, change the positions of the flow sender and the receiver, and therefore change the flow layout.Our objective is to reduce the total migration time and VDC downtime by determining a reasonable multiple VMs migration scheduling, thereby minimize the impact on cloud data center network and eventually achieve the purpose of traffic optimization. The details as follows:Firstly, we model the multiple VMs migration and conduct simulation experiment with one-by-one greedy algorithm. Results and analysis show that the differences are little between different greedy algorithms with single influence factor.Secondly, we design a multiple VMs migration framework based on SDN, which includes the centralized network controller and pre-defined forwarding rules. In charge of forwarding pre-defined rules in a source-to-destination fashion, this scheme can provide a flexible way to migration traffic via multiple routing paths at the same time.Thirdly, we propose an adaptive VM migration scheduler in multiple virtual data centers, named AVMS. The migration performance of multiple VMs results in generation an optimal order with consideration of fine-grained factors. To the best of our knowledge, this topic has not been widely studied in the literature. Furthermore, three scenarios are considered to evaluate the simulation experiments performance. It is demonstrated that AVMS can decrease the total migration time by up to 30.7% and average VDC downtime by up to 23.1%, compared with CQNCR algorithm.Finally, we design and build a VM migration demonstration platform based on Xen and complete the development of the relevant modules.