一种基于多核CPU的NDN数据名查找加速技术
首发时间:2018-03-28
摘要:命名数据网络(Named Data Network, NDN)是一种新型的网络架构,其基于数据的名称进行包转发。数据名不同于IP地址,具有长度不定、构成复杂等特点,因此其前缀匹配的查找时间空间效率相比IP地址更难得到保证。目前已有的研究方案中,有时间性能较好的二分查找方法LNPM(Longest Name prefix Match),但其存储开销较大;也有在LNPM基础上采用布鲁姆过滤器进行存储优化的方法BBS(Bloom-filter assisted Binary Search ),但是其在查找时间效率上有所牺牲。因此,本文在BBS基础之上设计了一种基于多核CPU的加速方法MBBS(Multi-core BBS),在保持低存储开销的前提下,也降低了最长前缀匹配查找的时间开销。实验证明,优化后的方法在吞吐量和单包延时方面都优于BBS,并且在CPU处于高负载情况下具有更好的稳定性。
For information in English, please click here
An NDN data name lookup accelaration technology based on multiple-core CPU
Abstract:Named Data Network(NDN) is a novel network achitecture, and it is based on name of data to forward packets. Unlike IP address, name has characteristics of unlimited length and complex compostion, therefore it is harder than IP address to ensure the time and space efficiency of lookup. In current researches, a approach named LNPM(Longest Name Prefix Match)based on binary search has great time efficiency but high memory cost, and anthoer approach , named BBS(Bloom-filter assisted Binary Search ), based on LNPM uses Bloom-Filter to reduce memory cost with slightly worse time efficiency. Therefore , an optimized approachbased on BBS, named MBBS(Multi-core BBS), is designed in this paper to acceleratelookup with multiple-core CPU, that keeps low memory cost and reduces time cost. Evalution show thatoptimized approach is better than BBS in terms of throughput and average packet delay, and has better stability when CPU is under heavy load.
Keywords: Computer Network NDN Name Lookup MBBS
基金:
引用
No.****
同行评议
勘误表
一种基于多核CPU的NDN数据名查找加速技术
评论
全部评论0/1000