摘要
根据IXP28XX网络处理器多引擎并行处理机制,针对微引擎处理数据包中出现的负载过重和任务拥塞问题,通过定义微引擎负载偏差率以及适应度函数,建立解决负载均衡问题的模型,利用遗传算法可以完成对各微引擎的负载调度。提出一种多点交叉和变异的遗传算法,仿真结果表明,该算法可以减少最优值迭代次数,从而缩短调度时间,使各微引擎的负载快速达到均衡。
According to the parallel processing of multi-micro engines in IXP28XX network processor, through defining the load deviation and fitness function, a model is built to solve the problems of over-load and congestion in dealing with the packets. Genetic algorithm is used to arrange the workload of every micro-engine in this model. A genetic algorithm with method of multipoint cross and mutation is put forward. The experiments show that this genetic algorithm reduces the iterations and saves the time in searching the optimum value, so the micro-engines quickly reaches the workload balance.
出处
《计算机工程》
CAS
CSCD
北大核心
2008年第5期101-103,共3页
Computer Engineering
基金
粤港关键领域重点突破招标项目(2005A10307002)
广东省工业攻关项目(2006A1010200)
关键词
网络处理器
遗传算法
负载均衡
network processor
genetic algorithm
load balancing
