基于遗传算法的网络处理器任务自动分配技术  

Genetic Algorithm-based Automated Task Allocation Technique for Network Processors

作　　者：王灵敏[1] 闫守孟[1] 周兴社[1] 谷建华[1] 

机构地区：[1]西北工业大学计算机学院,西安710072

出　　处：《计算机工程》2005年第22期186-188,共3页Computer Engineering

基　　金：国家"863"计划基金项目资助(2003AA1Z2100)

摘　　要：如何以提高系统吞吐率为目标,高效地实现报文处理任务集在NP处理引擎组上的映射与分配,是基于NP的网络应用系统设计的关键问题之一。在NP应用初期,任务分配主要以人工方式完成,易出错且开发效率低。经过深入分析,认识到NP任务分配问题具有多极值搜索特性,进而提出一种基于改进遗传算法的自动分配方法。实验证明该方法能较快得到接近最优的分配结果,可有效提高此类系统的开发效率。A problem that must be faced when designing systems based on network processors （NPs） is how to assign the packet-processing tasks to the processor pipeline consisting of many processing engines. Nowadays, this work is done by developers manually, which is error-prone and has low efficiency. To alleviate this, this paper proposes a genetic algorithm-based method to assign processing tasks to processing engines automatically. In essence, this is a multi-objective searching problem, thus the introduction of genetic algorithm is reasonable. Experiments show that this method can quickly obtain near-optimal solutions from the large solution space. Using this technique, the assignment of task to PE can be fulfilled easily and efficiently. Thus, the development period can be shortened greatly.

关 键 词：网络处理器 流水结构 任务分配 遗传算法 

分 类 号：TP393.02[自动化与计算机技术—计算机应用技术]