Research And Implementation Of Data Center Network Traffic Scheduling Strategy Based On SDN

With the repaid development of technologies such as big data and cloud computing,the traffic in Data Center Network(DCN)has grown up rapidly in recent years.How to schedule traffic in data center network to improve network performance and quality of services have become a hot research topic.Due to the lack of a global view of the network,traditional traffic scheduling methods may map multiple elephant flows to the same path,resulting in network congestion and hurt the network performance.SDN(Software Defined Networking)is a new network architecture that decouples the control plane and the data plane of the network,it can provide centralized management and fine-grained flow control capability for the distributed network and can better solve the problems exiting in traditional traffic scheduling methods.Therefore,this paper mainly focuses on the problem of traffic scheduling in data center network and studies the traffic scheduling strategy of data center network by combining with SDN technology,so as to improve the performance and quality of services of data center network,the main contents of this paper are as follows:1)Aiming at the shortcomings of traditional static traffic scheduling method may lead to poor quality of services of network when the network load increases because it does not consider links status and traffic characteristics.A dynamic traffic scheduling strategy called DSFlows(Dynamic Scheduling Flows)for quality of services guarantee is proposed.According to the network state obtained by the controller,the method can formulate the routing strategy for the traffic entering the network by comprehensively considering the hops of paths,the available bandwidth of paths,the delay of paths and bandwidth equalization each paths.The simulation results show that when the network load increases to 0.9,DSFlows increases the network throughput by 16.08%,and decreases the average round-trip delay and packet loss rate by 49.76% and 56.5% respectively compared with ECMP.2)In view of the unpredictability and high dynamic characteristics of data center network traffic,an EFRM(Elephant Flows Rescheduling Mechanism)for congestion awareness is proposed to further improve network performance by introducing traffic monitoring and traffic rescheduling.In addition,aiming at the high overhead of the controller after introducing traffic monitoring,the flow statistics collection method based on aggregationswitches and the adaptive polling period adjustment algorithm are proposed to reduce the message interaction between the controller and switches,and reduce the overhead of the controller for frequent analysis of network statistics.The simulation results show that when the network load increases to 0.9,DSFlows(+EFRM)increases the network throughput by22.24%,5.3 and 1.73% respectively compared with ECMP,DSFlows and Ashman,and it decreases the round-trip delay by 54.63%,9.68% and 31.49% and decreases the packet loss rate by 59.87%,7.75% and 26.69% respectively compared with ECMP,DSFlows and Ashman.Besides,compared with Ashman and Poll All Switches,the method can reduce the message overhead of the controller effectively when the network load is light.