Study And Implementation Of The Technology Of Multi-table Optimization On Soft Forwarding Plane

With the development of network technology,it has generated some new design concepts and new technological such as,Software Defined Network,(SDN)[1] and Network Function Virtualization(NFV)[2],Router,as an important part of the network infrastructure,how to ensure a high forwarding performance deployed on common platform implemented in software is the current research focus on.In the routing device,the greatest impact on performance is FIB(Forwarding Information Base)lookup and ACL(Access Control List)lookup,the two lookup tasks: FIB lookup is used to determine packet forwarding behavior;ACL lookup is based on the security control strategy of packets classification,that ACL,such as allowing or blocking,according to a predetermined strategy execution mode determine the packet.The speed of lookup two tables will determine the software router performance,after hitting part forwarding rules can only be queries a few ACL rules.Focused on the research of the optimization of router forwarding plane lookup task,analyzing the relevance of FIB and ACL,find rules of two tables can influence each other through the destination IP prefix.Therefore,we propose a dual-path forwarding plane lookup architecture(Dual-Path Lookup Architecture,DPLA),in the forwarding plane includes two data paths,the path FIB to Fast-ACL and the path FIB to ACL.In FIB designed two kinds of rules,overlapping and non-overlapping forwarding rules,which to determine which one packets path should forwarding.In DPLA,FIB associated with the ACL to each other,and the type of forwarding rules is a relative concept related to the position,so the type of forwarding rules will change when two tables update.Aimed at the two tables updated the effects on the forwarding rule type,this paper presents a method to calculate forwarding rule type changes in the rules of updates.In computing,fully consider the various possible situations,which to ensure the correctness of the forwarding rule type.In order to further validate the performance advantages of DPLA architecture in optimizing soft forwarding plane table lookup,DPLA was designed and implemented using general multi-core platform,including two parts the implementation of control plane computation and the implementation of data plane table lookup: the former is responsible for the rule classification and sent,the latter determine the packet forwarding behavior,interact between the two planes through an open control protocol implementation.In the experimental platform,FIB lookup using DIR-24-8 algorithm,ACL search using several typical algorithms(such as the hierarchical trie,expand trie,Hicuts,RFC algorithm).On this basis,through the experimental results show,when the data plane using different ACL searching algorithm,DPLA architecture is effectively,can improve data plane look-up table speed.Such as Hicuts algorithm can improve 18%,expand trie algorithm can improve performance by 20%.