For the issue that the task scheduling problem in heterogeneous Hadoop cloud computing platform,this paper analyzed the defects of speculative execution scheduling and LATE scheduling scheme,and proposed a task scheduling scheme based on task progress aware.Firstly,it estimated the approximate task completion time (ATE) according to the task progress on computing nodes,so as to determine the straggler task.Then,it divided the node into slow and fast nodes by the threshold value,which was 25％ of average task progress.When the straggler backup task reached a certain threshold,it would be assigned to the fast node priority.The experimental results show that the proposed scheme can reasonably schedule tasks for heterogeneous Hadoop platforms,and effectively reduces the task completion time,response delay,and improves the throughput of the platform.%针对异构Hadoop云计算平台中任务调度问题进行了研究,并分析了现有推测执行调度和LATE调度方案的缺陷,提出一种基于任务进度感知的自适应任务调度方案.首先,根据当前计算节点上的任务进度情况,估计任务近似完成时间(ATE),以此确定掉队者(straggler)任务;然后,以平均节点任务进度的25％为阈值,将节点分为慢节点和快节点.当straggler后备任务达到一定阈值时,将其优先分配到快节点中执行.实验结果表明,提出的方案能够为异构Hadoop平台合理分配任务,有效降低了任务完成时间和响应延迟,同时提高了平台吞吐量.
展开▼
