Time-Space Combined Traffic Scheduling Mechanism For Data Center Network Based On SDN

With the next-generation network technologies on the fast increase and a great quantity of distributed services spring up,data center network is broadening continuously,and network traffic is more complex.Software defined network(SDN)separates the control function from the data forwarding function.It can achieve software implementation of elastic management,which is well applied in data center networks.However,the SDN-based traffic scheduling schemes available are mostly aimed at the scheduling process of a single flow,and lack of description of the overall features and the service characteristics of traffic.Therefore,how to use the superiority of SDN centralized control to more efficiently and specifically schedule and manage the traffic of data center network is a problem that needs to be considered.This thesis is devoted to the research on the overall features and service characteristics of the flow in the data center network.SDN-based traffic scheduling schemes are respectively designed from the perspective of space and time-space combination,in an attempt to elevate global performance and conserve network resources.The main contributions of this thesis are as follows:Based on the analysis of the overall features of flows,a scheduling scheme on the basis of fruit-fly optimization algorithm is designed from the perspective of space.The concepts of traffic aggregation and subgroup are put forward.Many flows of the same origin and same sink are get together uniformly.The average load balancing degree of the distribution paths of multiple aggregated flows is taken as the optimization objective.The fruit-fly optimization algorithm is brought and deployed in the controller to realize the simultaneous optimal scheduling of multiple flows.The advantage of this scheme is that for multiple flows with the same origin and destination,it is not necessary to perform scheduling arrangements through the controller one by one,but is scheduled by the controller uniformly at one time,which can reduce the transmission completion time of flows on the whole,conserve network resources and control overhead.Simulations indicate that the scheme can promote average throughput,load balancing degree and link bandwidth utilization.Based on the analysis of the service characteristics of flows,a scheduling scheme based on service categories and priority is designed from the perspective of time-space combination.In terms of time,quantitatively set the priority of traffic with different service categories,and perform scheduling in the light of the priority order.In terms of space,different optimization goals and scheduling algorithms are assigned to the traffic with different service categories to select the optimal routing.Finally,time and space,priority and routing are combined to schedule and process the traffic with different service types.Simulations indicate that the scheme owns better properties in average throughput,average round-trip delay of delay-sensitive flows and packet loss rate.