摘要
对常规的服务器集群架构进行了改进,提出了决策器的概念,并由决策器训练调度序列;同时根据自适应小生境遗传算法提出了适应特征值作为适应值的评价标准,并合理地应用于负载均衡调度系统中.本文搭建网络环境模拟大并发测试,实验数据与分析表明,本方案的系统平均响应时间仅为2ms,同时错误率趋近于0,相比改进前服务器更均衡地利用,系统性能更稳定.
This article has made improvement in the architecture of regular server clusters, The concept of ＂decision maker＂ is put forward, which trains scheduling sequence; Moreover, according to adaptive niche genetic algorithm （ANGA）, the adaptive eigenvalue is regarded as evaluation criterion which can be reasonably applied in the controlling system of load balance. Furthermore, a network environment is simulated by thousands of test. It is demonstrated by experimental data and analysis that the average responding time of the system is 2 ms. At the same time, error rate approaches 0. Compared with the system before improvement, the new system is more balanced use and stable in performance.
出处
《微电子学与计算机》
CSCD
北大核心
2013年第12期54-56,60,共4页
Microelectronics & Computer
基金
铁道部重大项目(2012X004-A)
四川省科技支撑项目(2011RZ0003)
关键词
负载均衡
小生境遗传算法
动态轮询
. load balance~ adaptive niche genetic algorithm （ANGA） ~ dynamic round robin（drr）
