Cloud Task Scheduling Algorithm Based on Three Queues and Dynamic Priority
Yanyue Yu, Yu T. Su · 2019 IEEE International Conference on Power, Intelligent Computing and Systems (ICPICS) · 2019
In view of the current situation that the heterogeneity of cloud computing cluster is increasing and the load difference of system is obvious, this paper proposes a cloud task scheduling TQ (Three Queues) algorithm based on three queues and dynamic priority. TQ algorithm first puts jobs into the waiting queue according to the priority of jobs, and then divides jobs into job types according to the data input amount of Map phase, data output amount of Map phase, total number of current node running tasks, completion time of Map tasks and disk I/O rate. Jobs are put into corresponding queues to improve hardware utilization. The experimental results show that the algorithm can effectively improve the performance of cloud task scheduling under the coexistence of I/O-intensive jobs and CPU-intensive jobs, and shorten the completion time of the total task.