Reasearch Of Traffic Optimization Strategy In Data Center Network Based On SDN

With the emergence of the concept of “cloud”,4K、Big Data and Internet of Things flourish,the appearance of all these new technologies are related to “cloud”.Those all lead to a rapid growth trend of the traffic in the core infrastructure of “cloud” ——datacenter networks.In the face of greater traffic size,more application deployment,higher quality of service requirements,how to solve the problem of datacenter traffic optimization,such as improve network throughput,reduce network latency,and achieve flexible scheduling,has become an urgent problem to be solved.The emergence of Software Defined Network(SDN),represented by the Open Flow protocol,has brought new opportunities for the resolution of the problem.In datacenter networks,The ununiformity of traffic is one of the reasons leading to the blocking of network.The routing scheme that balances network load amongst multiple available paths can reduce blocking.Nevertheless,the state-of-the-art multipath routing schemes use static hashing to route data flow to different paths,which would cause blocking if the traffic belongs to elephant flows.Besides,the flow-separation approach makes the packets out-of-order.Aiming at solving the above problems,this paper takes advantage of SDN and the characteristics of different transmission requirements of mice and elephant flows,and proposes a load balancing mechanism named IRS(Integrate Routing and Scheduling)for datecenter networks.In order to distribute traffic uniformly in network,IRS adopts greedy routing and multiple paths routing for elephant flow transmission.Besides,IRS introduces packet scheduling based on ip addressing to avoid packets out-of-order.The experimental results indicates that IRS is 12% better than ECMP in load balancing performance.A datacenter network is generally a multi-tenant network environment.Creating multiple tenant networks on a single physical network,providing their own network views for different tenants,and realizing that traffic does not affect have become the urgent need for datacenter network development.But in the datacenter networks based on SDN,limited flow table resources put forward new requirements for the management of tenant network traffic,Because of the characteristics of the irregularity of the topology and the randomness of the flow,Unreasonable traffic management mechanism will lead to the ununiformity of flow table using in switches,which may lead to flow table overflow phenomenon at the network bottlenecks,an increase in SDN controller load and a decrease in network performance.This paper designs a network resource allocation and management platform based on SDN for multi-tenant,and verifies the the resource allocation function and traffic isolation function of the platform.And,this paper also proposes a tenant network traffic management scheme which simultaneously balances flow table using and the network load,then the performance comparison experiment is carried out on the platform designed in this paper.The experimental results indicates that the proposed scheme reduces the transmission delay compared with the scheme just based on the residual flow chart and effectively balances the use of the flow table resources compared with the scheme just based on the routing of the remaining bandwidth.