Reaserch On Server Load Balancing Based On Openflow

Data centers have become the critical Internet infrastructures, undertaking tasks of data storage, computation and transmission. Many kinds of critical business systems are located within data centers, such as search service, distributed databases and web services. Rich applications and services make the scale of data centers expand. It brings a great challenge to servers and transmission capacity of network because of complex data processing services and the mode of cluster communications within data centers. Only using expensive, high-performance servers and switches is likely to cause a waste of system network resources, and increase the maintenance costs of data centers without reasonable resource management strategies. To a certain extent, virtualization technology can solve the above problems. By using virtual machines can take full advantage of physical resources on servers and balance load among servers. However, the migration of virtual machines will also change the transmission path of related network traffic at the same time. The changes of distribution of network traffic will make it easier to cause link congestion, packet loss, and the long transmission delay, thereby affecting the performance of the applications running on the virtual machines. So how to migrate virtual machines to ensure the server load balancing while avoiding the network congestion and optimizing the communication delay between servers has become a serious problem.This paper mainly focuses on server load balancing policies in the data center based on OpenFlow network. OpenFlow protocol supports the management and configuration of network equipment and data flows, and can monitor the status of network. The centralized controller of OpenFlow network makes up for the disadvantages of traditional distributed network. This paper first introduces the principle of virtual machine migration technology, the characteristics of the architecture of software defined network as well as the development of SDN and OpenFlow protocol. Then we design the framework of server load balancing in the data center to monitor the resources usage of physical servers and virtual machines. Getting the network topology and the available link bandwidth through OpenFlow controller as the basis of server load balancing algorithm. A server load balancing algorithm is designed to migrate virtual machines to balance load among servers. Not only the utilization of system resources, but also the communication cost according to the transmission delay or network bandwidth, demanded by the applications running on the migrated virtual machines, is taken into account by the algorithm to avoid affecting the quality of service due to network congestion or limitations of system resources. Finally, the server load balancing framework and the algorithm are tested on Mininet, the simulation platform of software defined network, to validate the effectiveness of the framework and algorithm. Then we compare the algorithm with the greedy strategy only considering meeting the system resources requirements of the migrated virtual machines.