Research On Data Center Multipath Traffic Scheduling Strategy Based On SDN

The single management method and fixed operation mode of the traditional network have been unable to complete the scheduling of internal traffic of the data center with flexibility and high quality,so how to solve this problem has become a research hotspot.The emergence of Software Defined Network(SDN)provide new ideas for traffic transmission in the network.It utilizes the characteristics of control and forwarding separation and the centralized control,the problem of data center traffic scheduling is well solved.Based on the software-defined network architecture,this thesis studies the traffic scheduling problem of the data center.The main contents are as follows:1.Aiming at the shortcoming that SDN controller only considers the link load load to calculate the optimal path of data flow,a traffic scheduling strategy based on multi-factor is proposed.The proposed strategy combines parameters such as the number of large flows,bandwidth,and delay to calculate the optimal path for the traffic.Firstly,the K-shortest path algorithm is used to calculate the path set with the least number of hops from the source to the destination,and then in order to more accurately measure link-state,according to the traffic characteristics of data center,introduces the link criticality which related to the number of large flows and the sum of link traffic,and finally introduce the link cost composed of the link bandwidth and the delay of the weight value determine the optional path.Simulation results show that compared with ECMP,Hedera and PureSDN,the proposed strategy has a better effect on improving network throughput and link bandwidth utilization performance indicators.2.A dynamic probability jump traffic scheduling strategy based on link load is proposed for the local congestion problem caused by large flows with large bandwidth.The proposed strategy periodically monitors the link load information of the network based on the shortest path set of the source host to the destination host,and then designs the path probability jump type according to the real-time load of the link,at the same time,in order to make full use of network resoures,the jump threshold is designed,the last traffic selects different routing methods according to the probability jump value of the path.When a path becomes busy,that is,when the path jump probability value is greater than the jump threshold,the flow with the largest load on the path will jump to another path with a lighter load;if it is less than the jump threshold,The route with the maximum probability value is selected for the new traffic.Simulation results show that compared with ECMP and Hedera,the proposed strategy has better performance in improving the average throughput of the network,reducing the average round-trip delay of the traffic,and reducing the average deviation of the round-trip delay of the traffic.