Research And Design On Routing Strategy Of Data Center Based On OpenFlow

After decades of development, Internet has become an indispensable part of our life. In recent years, with the development of cloud computing technology and the proposed concept of Big Data, a lot of new types of Internet applications have emerged. Therefore, the data center occupies a pivotal position in the Internet. Today, the Internet, including the data centers, have face many problems. Traditional architecture of the Internet has been unable to satisfy the new requirement of the new emerging network services and applications. The appearance of SDN architecture and the OpenFlow technology enables the network programmability by decoupling the control plane and the forwarding plane of the network and provides a new way of solving the problems faced by data centers or even by the Internet.In this paper, we use OpenFlow as our main technology, and focus on the routing strategies used in the data center network, to solve the traffic load problem. Most of the existing routing strategies used in the data center network are static ways, which lack of the feedback information from the network, resulting in irrational distribution of bandwidth resources in the network and not being able to take full advantages of the redundant bandwidth resources provided by the data center, making a waste of resources.Firstly, we propose a flow classification-based dynamic routing strategy for data center network. The strategy monitors and collects the links’and flows’information in the real time by using the OpenFlow technology. These collected information is used as reference to calculate the route for the flows in the network. For the new coming flows in the network, we take the greedy principle to choose their route with the link information in order to achieve load balancing purpose. Then, by using the collected flow information, we classify the flows being forwarded in the network into small flows and big flows, and reroute for the big flows dynamically to make the big flows getting more bandwidth, improving the utilization of the links and reducing the probability of congestion.Secondly, we also propose a congested routing management policy based on the OpenFlow technology. By real-time monitoring of the status of links, when a congested link detected, the congested routing management policy starts to work and reroute the big flows on this link to find better routes for them. When rerouting these flows, we take the way back from the starting point of the congested link, thereby reducing the cost of rerouting. This congested routing management policy is triggered after the congestion occurs and works until the congestion state ends. It is a passive way to relieve the congestion occurred in the network. The congested routing management policy is the complementary for the flow classification-based dynamic routing strategy.Finally, our implement uses Mininet simulation platform. Compared with existing routing methods, the evaluation results demonstrates that our strategies are superior over other policies which can improve the link utilization and the throughput of the network.