桌面网格中面向资源可用性预测的任务调度算法

提出桌面网格平台下的一种面向资源可用性预测的任务调度算法.该算法充分考虑了计算资源在执行作业的过程中可能发生的行为,采用预测技术保证了任务的高效而合理的分配.当计算资源发生异常时,通过公平的转移权重预测方法估计资源在下一阶段可能的状态,计算出资源的可靠性概率,然后开始调度子任务给资源.通过建立实验环境,设置不同的可靠性域值T与历史检查资源天数N等参数,在桌面网格上进行了测试.最后把该调度算法的实验结果与PPS等调度策略进行比较,验证了本文的任务调度算法在子任务处理率与通信轮回时间上有比较好的性能.

Resource Availability Prediction Oriented Task Scheduling Algorithm in Desktop Grid Platform

A Task scheduling algorithm which is based on an existing resource availability prediction technique in Desktop Grid was proposed and described.This approach was to recognize what will happen to the computing resources and make reasonable job allocation decisions accordingly.It uses a resource′s availability history to predict the probabilities of that resource next exiting the available state into each of the non-available states,to calculate these probabilities;a Transitional NDay with Equal transition weights(TDE) was also used.The prediction technique calculates the output probabilities by counting both the number of transitions from available to other states and how many times the job could be processed between two transitions.We used simulation to validate our approach and show the subtask process rate always increases when the reliability threshold T added.The results also indicate that checking more days does not provide better prediction performance for subtask scheduling.