Designing And Implementation Of Load Balancing Methods In Data Center Networks Based On OpenFlow

With the rapid development of next-generation IT technologies such as cloud computing and big data,data center networks are facing tremendous data traffic pressures,and traditional load balancing technologies cannot meet their performance requirements.As a new type of network architecture,Software Defined Networks(SDN)has advantages of the separation of data plane and controller plane,global centralized control,and is applied to data center networks increasingly.This article focuses on load balancing methods in the data center network,with the advantage of the global centralized control of the SDN architecture and a global resource view is established,the article build an SDN simulation platform,design and implement a load balancing algorithm,clarify and improve the rerouting mechanism of the OpenFlow protocol.It improves the load balance of data center network effectively and improves the efficiency of data forwarding.The main work is as follows:(1)The cost of building a real SDN network platform is too expensive.At the same time,there is a shortage of directly available SDN simulation software.This paper firstly analyzes the requirement of SDN based simulation platform in data center network,and then builds an OpenFlow based SDN simulation platform,The data forwarding layer,control layer,data storage layer,data visualization layer and other functions are implemented.Finally,simulation experiments about the Floyd algorithm and the ECMP algorithm are carried out on the simulation platform to verify the usability of the SDN simulation platform,which provides a basic condition for subsequent algorithm and mechanism research.(2)Aiming at the problems such as unreasonable resource allocation and low resource utilization in current data center networks,a Multipath Transmission based Dynamic Load-balanced Routing(MTDLR)algorithm is proposed and implemented.The algorithm defines the path selection function considering the path level network status such as optimality of the average link available bandwidth of the path,uniformity of the link available bandwidth of the path and the hops of the path,and adjusts the weights of the influencing factors in the path selection function to adapt to different network topology requirements.Based on the SDN simulation platform constructed in Chapter two,the Poisson traffic generation module and the MTDLR algorithm module are added.The MTDLR algorithm is simulated.The simulation results show that,compared with the GLB and ECMP algorithms,the MTDLR algorithm has a certain degree of performance improvement in terms of average flow bandwidth utilization,system throughput,average packet loss rate and average delay jitter,which can achieve the goal of balancing Network load and improve network resource utilization.(3)Due to the conditions of constantly network changing during the data transmission,rerouting is introduced to further improve network load balancing performance,and a Traffic Monitoring with Dynamic Rerouting supported(TMDR)mechanism is designed and implemented which includes path monitoring mechanism,congestion monitoring mechanism,scheduling selection mechanism,rerouting mechanism and monitoring removal mechanism.The paper defines two indicators of path bandwidth utilization and flow equilibrium degree to identify network congestion and trigger dynamic traffic scheduling.On the basis of the SDN simulation platform built in chapter two,the initial routing module and the TMDR mechanism module are added,and the TMDR mechanism is combined with MTDLR algorithm and GLB algorithm respectively to simulate the experiment.The simulation results show that after the introduction of the TMDR mechanism,the average flow bandwidth utilization,system throughput,average packet loss rate,and average delay jitter performance are further improved.