摘要： 从网络处理器的体系结构入手，在线程和微引擎层次上对IXP系列网络处理器的并行环境进行抽象分析，在线程方面阐述了顺序和非顺序情况下不同的编程方法及其对应的效率和吞吐量，在微引擎方面以获得最大的吞吐量为目标给出最优的微引擎分配模型。针对IPv4路由器这种应用进行求解，当其采用1rx3pr1qm1sc2tx的分配方案时系统吞吐量最大，达到4 670.47 Mbp/s。
关键词:
网络处理器,
线程调度,
微引擎分配,
性能分析
Abstract: This paper makes an abstract analysis on the parallel environment of IXP series Network Processors(NP). This work is done at both ME and context levels. As to contexts, different programming approaches, for sequential and non-sequential cases, are explored for efficiency and throughput analysis. At ME level, an optimum ME allocation model is presented with an objective of maximum throughput. Solutions are given to a specific application of IPv4 router, and the system reaches the maximum throughput 4 670.47 Mbp/s when the distribution scheme of 1rx3pr1qm1sc2tx is adopted.
Key words:
Network Processor(NP),
context scheduling,
ME allocation,
performance analysis
中图分类号:
魏 帅;赵荣彩;单 征;毛 艳. 基于IXP系列网络处理器的并行原理及其应用[J]. 计算机工程, 2008, 34(12): 102-104.
WEI Shuai; ZHAO Rong-cai; SHAN Zheng; MAO Yan. Principle and Application of Parallelism for IXP Series Network Processors[J]. Computer Engineering, 2008, 34(12): 102-104.