Research On Multi Path Load Balancing And Flowtable Distribution Optimization Based On SDN

With the development of computer networks and the emergence of big data and cloud computing,the data traffic in the network growth rapidly,Software Defined Network(SDN)and data center network has become one of the main directions of the future network development.SDN defines a new network architecture,the core idea is to separate the control plane from the data plane,realized by controller and network equipment.Different from the traditional IP network,the centralized control of SDN network are all implemented by the controller,while network equipment is only used for high-speed data forwarding.The data center network can solve the problems that caused by the increasing network traffic through the centralized storage of network data information to realize the information centric network architecture.The researches on SDN network and data center network has been drew more and more attention by researchers.In order.to improve the network bandwidth and throughput,datacenter network employ layering network topology structure with multi-path characteristics,such as Fat-tree.However,traditional routing algorithms have limited support for multi-path routing.In order to solve this problem,the SDN is introduced into the data center network,and the network resource can be fully utilized to realize the multi path load balancing by programming the control plane.In SDN network,controller manages forwarding behaviors of the network devices through standardized OpenFlow protocol.After controller formulates the routing policy,it sends flowtable to the OpenFlow switch,and data streams is forwarded by the matching flowtable item.Flowtables of the OpenFlow switches are implemented by Ternary Content Addressable Memory(TCAM),with only have about 1k-2k flowtable entries,flowtable resource are limited.When the data flow does not match the flowtable entry,it will sent a packet-in message to SDN controller to request flowtable resource.Controller can handle only about 30k packet-in messages per second,computing resources are limited.Therefore,it is necessary to balance the use of flowtable resources and computing resources to optimize flowtable allocation plan.The main contents of this paper are as follows:(1)In order to solve the problems that the traditional routing algorithm has limited support for multi path in Fat-tree data center network,this paper proposes a multi path load balancing algorithm(F-MPLB).The algorithm uses the characteristics of SDN network centralized control to obtain multiple paths real-time status information,evaluate the path bottleneck bandwidth based on the status of switch node and link load,in order to choose the best path to meet the requirements of data flow transmission.The experiment results show that F-MPLB algorithm is superior to traditional routing algorithms in not only reducing network propagation delay but improving network throughput.(2)For the limited switches flowtable resources and SDN controller computing resources,this paper provides a resource preference flowtable distribution optimization algorithm(RC-FDO)in SDN network,the algorithm study on the characteristics how to set reasonable flowtable idle_timeout value,combined with the current network running status and data flow,by using concepts of flowtable resource cost,calculation resources cost and resource preference.The experiment results show that RC-FDO algorithm can effectively improve the rate of table resource utilization and data flow acceptance,to achieve a higher average throughput of the network.