Scheduling algorithm for instance-intensive cloud workflow
Kaigui Wu · Journal of Computer Applications · 2010
The existing workflow scheduling algorithms are simply designed for single complex instance, unsuitable for scheduling instance-intensive cloud workflows. To address this problem, a new scheduling algorithm, named Minimum Total Cost Under User-designated Total Deadline (MCUD), was proposed based on multiple instances. For the workflow instances of the same type, after classification, MCUD algorithm distributed the user-designated overall deadline into each task with a new distribution method. In addition, MCUD algorithm adjusted the sub-deadline of successive tasks dynamically during the scheduling process. Instances of the same nature are given the sub-deadline distribution results of some difference, which can avoid the fierce competition of cheaper services and increase the efficiency of resource utilization. The simulation results show that MCUD algorithm further decreases the total execution cost and total execution time while meeting the user-designated deadline in comparison with other algorithms.