영문내용
(English Abstract) |
In a cloud environment, users can request resources such as virtual machines and perform complex calculations. With many services and resources available, execution time for each application can satisfy the deadline. By implementing scientific workflows on the cloud with scalable and flexible resources, this method reduces the amount of costs for execution. This paper proposes a mechanism to determine what virtual machine instances used by each task, by splitting the workflow into levels, and determine the task with the highest completion time and adjust it by changing the virtual machine assignment. A dynamic resource re-provisioning and task rescheduling in dealing with possible delays that are present in real life systems. Simulation results showed that implementing our approach fitted the deadline constraint, as well as providing efficient results compared to the baseline algorithms.
|
