中文摘要：

考虑到网格中存在着大量功能相同但可靠性各不相同的资源,针对执行时间限制严格类型的网格工作流提出了一种新的基于状态预测的资源选择策略。该策略以工作流DAG图中的关键路径上的前驱任务资源收到输入转移数据而被触发执行的时刻作为预测时刻,利用Markov过程计算出后继任务的各侯选资源在前驱任务结束时刻处于＂闲状态＂的概率大小,然后选择这一概率较大的资源作为此后继任务的执行资源,从而了保证了工作流关键路径任务执行时间的要求。最后通过实验验证了该策略的有效性。

英文摘要：

Considering that the number of grid resource in grid environment is very likely more than one and that all of these resources have the same function except for their reliability, to the workflow ofDAG-based workflow in which time constraints are rigid, a new resource choice strategy based on the model of resource states prediction is put forward. Taking the time for the predicting time when the previous task ofworkflow on the critical path is triggered to execute by inputting data, the strategy obtained the probability for each of resource of the subsequence task being idle state at the end of the previous task executing through employing Markov process, and then made choice of the resource with larger probability for the subsequence workflow task to meet the deadline ofworkflow tasks on the critical path. The simulation shows the validity of strategy.