The Research On The Application Of SDN In Optimization Of ARP Traffic

Cloud computing technology is a comprehensive technology stack that is stacked on the basis of traditional network,storage and virtualization.It aims to change the way and utilization of traditional resources.Data center as the primary application of cloud computing,urgently need for such resources utilizaton pattern to improve the use of computing,storage and network resource utilization.In the traditional non-cloud data center,the client-server model is the main way that is north-south traffic;In today's cloud data center,big data as the representative of the distributed computing create the previous traffic model.In the cloud computing data center,the east-west network traffic model requires a lower latency between virtual machines.Combined with cloud architecture analysis and network virtualization,the demand for the virtual machine migration to the layer-2 brings to the traditional network Greater challenge.Because of the traditional layer-2 network inherent flaw,the layer-2 broadcast message flooding mechanism is difficult to deal with the cloud data center needs of the big layer-2.Big layer-2 technology is proposed to achieve cross-data center virtual machine migration;the global can be layer-2 communication,so we can guarantee data center data to be high availability.However,the traditional layer-2 network inherent flaw,result in the size of the second floor that can not be too large and will have a broadcast storm that will further reduce network performance,resulting in increased network fluctuations and delays.The traditional network-based broadcast message processing mechanism is the decisive factor limiting the development of the layer-2.A large number of broadcast messages flood with valuable network bandwidth resources,and broadcast storms in addition to occupy the network bandwidth,also consume dedicated network processor processing time and resources,resulting in greater network latency.Just because of the flooding of the two-layer broadcast message processing mechanism led to the LAN arp attack,the local area network arp attack use the the layer-2 broadcast design flaws in order to forge arp response to change the direction of network data traffic,analyze and steal a variety of data of the network users.Thus,flood-based broadcast message processing mechanisms have completely limited the development of cloud data centers.Software definition network(SDN)as a gradually mature network technology,gradually demonstrate its advantages,especially in the data center applications.Global control of the information and logical centralized control is the core design ideas;the controller with a certain standard define the underlying forwarding equipment and its control functions abstract control functions to the controller side;its northbound interface can quickly define,deploy and adjust the network configuration.This paper designs a mechanism to avoid the flooding of broadcast messages in the cloud of cloud data center according to the big layer-2 of cloud data center,the network virtualization,the flooding of broadcast messages and the global information control and logical concentration of software definition network.The virtual network of the platform changes the initialized information and intercepts the arp request,and uses the SDN controller's global information control capability to send the arp request to the SDN controller according to the SDN OpenFlow communication protocol in the cloud virtual network and utilize the SDN controller framework Provide the Northbound API interface to design custom arp processing mechanism and to forge arp response.This design avoids flooding of broadcast packets in the network and effectively reduce the network bandwidth load and communication delay.Combined with the current data center business network traffic characteristics and OpenStack cloud network architecture,the paper design the cloud network initialization and specific protocol data direction.According to our design,based on the maximum compatibility of OpenStack's network at this stage,the minimum optimal adjustment is made to the architecture of the OpenStack cloud network's network traffic.In the second chapter of the paper,I mainly complete the virtual switch initialization changes.In addition to adding the default flow table,we also make the appropriate changes to match The purpose of the modification on the underlying virtual switch.According to our optimization design,we analyzed the working mechanism and architecture of OVS.In the third chapter of the paper,we analyze the architecture design and technical characteristics of the selected controller.According to the definition of the protocol data structure and the event dispatching mechanism,the ARP splitting mechanism is designed and implemented to improve the response speed of the controller.And we modify the controller framework to achieve the global processing of arp requests,so we achieve the OpenStack arp no flood direction processing.The fourth chapter is the verification and performance analysis of the function of the design.The optimization effect of the design is verified and compared with the parameters.The optimization effect of the design is proved.In short,the combination of SDN design ideas to achieve a non-flooding broadcast traffic handling mechanism make obvious change in the network performance.