摘要
提出一种基于遗传神经网络的主机负载预测模型，并基于该模型设计了集中式任务调度算法CJD—HLP。CJD—HLP采用预测法提前获得主机负载信息，保证了任务调度时使用决策信息的实时性、准确性，避免了负载迁移的抖动问题。实验结果表明，该算法较基于实测法的其他任务调度算法在性能上有较大提高。
A host load prediction model based on heredity-neural network is proposed, and center task scheduling algorithm CJD-HLP based on the model is designed. CJD-HLP algorithm adopts prediction method to get the host load information earlier, ensure that the information for decision-making is real-time and accurate and avoid the load transfer jitter problem. Experimental results demonstrate that the algorithm performance is greatly enhanced compared with the algorithm based on actual measurement.
作者
谢芳清
覃毅
XIE Fang-qing, TAN Yi （Zhongkai University of Agriculture and Engineering, Guangzhou 510225, China）
出处
《电脑知识与技术》
2010年第01Z期388-390,共3页
Computer Knowledge and Technology
关键词
负载预测
动态任务调度
网络并行计算
PVM
load prediction
dynamic task scheduling
network parallel computing
PVM
