改进的自动回归网格主机负载预测模型

网格计算中对资源的有效预测能很好的改进任务分配和作业调度的策略，提高它们的执行效率，作为网格资源预测的核心？主机负载的预测显得尤为重要。文中提出了一种基于AR改进的主机负载预测模型，它不仅具有AR模型本身的计算成本小、预测性能稳定的优点，还对AR模型只对未来某个固定时间段的负载预测进行了改进，使之能根据作业的预测执行时间进行主机负载动态预测，同时该改进模型还充分体现了主机负载变化的自相似性和长期依赖性，实验结果表明，该模型达到了预期的效果。

An Improved AR-based Model of Host Load Prediction in Grid Computing

Validity of resources prediction in the grid environment can better the strategy of task allocation and job scheduling,and increase the efficiency of their execution.As the main part of resource prediction in the grid computing, host load is particularly important.The paper presents an improved model AR-based of host load prediction.The improved model not only has the advantage what AR model has,such as,low computing cost and high stability,but also has the advantage that AR model has not,such as it can predict the dynamic interval host load when some job will encounter in the future,but the AR only can predict some fixed time in the future,on the other hand the model includes some character,for example,self-similarity and long-range dependence of the host load.Experiment result shows that the model comes to an anticipant result.