A Hadoop Dynamic Scheduling Strategy Based on the Node Performance
The default task scheduling algorithm of Hadoop can not adjust the assignment of task according to the performance. Proposes a dynamic task scheduling strategy based on the node performance. The strategy uses the node CPU utilization and memory utilization as the basis of evaluating the node's runtime performance, then adjust the task number dynamic according to the basis, so each node in a cluster can work in the good state. Experiments results indicate that the strategy makes the total completion time of task clusters reduced significantly and improve the performance of the cluster.
NETL
NSTL
万方数据
维普
