Research And Implementaton Of Traffic Scheduling Method Based On Fattree Data Center Network

With the development of services such as cloud computing and big data,the traditional data center network architecture has gradually exposed many problems,such as the low overall utilization rate of network links and long data transmission delays.In order to improve the throughput and fault tolerance of the data center network,researchers have proposed many new network topologies,the most classic of which is the Fat-Tree network structure studied in this thesis.However,when large and small flows coexist,the data transmission effect of the default ECMP routing algorithm in Fat-Tree topology is not ideal.Therefore,how to improve its internal routing scheme has always been a hot issue in the field of traffic engineering.The main research content of this thesis is as follows.(1)Designed traffic routing optimization methods for the large and small flows inside data center.Considering that the most important performance parameters of large flows in transmission are the transmission bandwidth and packet loss rate,this thesis designs a routing optimization method based on an improved ant colony algorithm for large flows.Considering that the most important performance parameters of small flows in transmission are delay and packet loss rate,and small flows have higher requirements for forwarding quickly.This thesis separately designs a fast routing method based on real-time network parameters for it.(2)A traffic routing optimization system that distinguishes large and small flows is designed,and the functions and implementation principles of its modules are introduced in detail.In order to realize the routing optimization method proposed for large and small flows,this thesis realizes the collection of real-time network status information based on SDN technology,provides necessary parameters for the routing method.After the routing decision completed,the modification of the traffic routing path is completed through the issuance of the flow entry.(3)The functions of each module of the traffic routing optimization system designed in this thesis are realized,the simulation environment required for system operation is built,and a variety of traffic sending modes are set to verify the operation effect of the system.The simulation results show that comparing ECMP and Hedera＿GFF routing methods under the same conditions,the traffic routing optimization method designed in this thesis improves the average throughput and average link utilization of the network while reducing the average transmission delay and average packet loss rate of small flows,improving the data transmission quality of both large and small flows at the same time.