| 引用本文: | 张晓哲,彭伟,朱培栋.基于并行流水的转发引擎设计与性能分析.[J].国防科技大学学报,2005,27(1):64-68.[点击复制] |
| ZHANG Xiaozhe,PENG Wei,ZHU Peidong.Parallel Pipeline Forwarding Engine Design and Performance Evaluation[J].Journal of National University of Defense Technology,2005,27(1):64-68[点击复制] |
|
| |
|
|
| 本文已被：浏览 7426次 下载 5959次 |
| 基于并行流水的转发引擎设计与性能分析 |
| 张晓哲, 彭伟, 朱培栋 |
|
(国防科技大学 计算机学院， 湖南 长沙 410073)
|
| 摘要: |
| 光通信技术对核心路由器报文转发能力不断提出更高的要求。10Gbps光传输技术已经使现有的各种软硬件路由查找方法成为核心路由器转发能力的瓶颈，而更高性能的光传输技术则已经突破了存储器访问速度的极限，使得基于单片存储器的路由转发方法无法应付未来日益增长的需求。在硬件存储器价格非常低的前提下，提出一种使用多个存储器并行流水查找的硬件转发实现结构。通过使用Internet上真实报文数据进行的性能模拟可以看出，随着并行度的增加，整个转发结构可以获得近似于线性的性能加速比。 |
| 关键词: 核心路由器 转发引擎 DRAM 模拟器 |
| DOI： |
| 投稿日期：2004-10-20 |
| 基金项目:国家自然科学基金资助项目(90204005)；国家863高技术资助项目(2003AA121510) |
|
| Parallel Pipeline Forwarding Engine Design and Performance Evaluation |
| ZHANG Xiaozhe, PENG Wei, ZHU Peidong |
|
(College of Computer，National Univ. of Defense Technology, Changsha 410073，China)
|
| Abstract: |
| With the rapid development of communication technology, optical transmit technique demands higher requirement of the core router's forwarding performance. OC192 POS interfaces have made the core router's forwarding engine a new bottleneck. Further development of optical transmit technique has exceeded the maximum accessing ability of DRAM, which makes impossible the design of new forwarding engine with one chip of DRAM. New hardware parallel forwarding engine design is offered on the basis of Gupat's DIR 24 8 BASIC forwarding architecture, taking advantage of the parallel of multiplex DRAM chips. In order to get really performance of the parallel forwarding engine design, IP packet header trace and routing table's dump of Internet core router node are used as the input of the target system's simulator. The result shows that the parallel forwarding engine can achieve linear speedup with the increase of parallel basic forwarding tables. |
| Keywords: core router forward engine DRAM simulator |
|
|
