An effective heuristic for due-date-constrained Bag-of-Tasks scheduling problem for total cost minimization on hybrid clouds

Bag-of-Tasks (BoT) applications consisting of multiple tasks widely exist in numerous fields. As customers use cloud resources in a pay-as-you-go way, they are willing to execute BoT applications on clouds. Cloud providers and customers establish contracts in which applications' due dates are specified. If an application cannot be finished before the due date, the cloud provider should pay a tardiness penalty. When the private cloud has insufficient available resources to afford all customer-submitted BoT applications, the cloud provider has to outsource some tasks to public clouds with resource-used costs. The key challenge here is how to schedule tasks on hybrid clouds to minimize the total cost, including all applications' tardiness penalties and the cost of using public clouds' resources. We study and formulate this problem as an Integer Programming. Accordingly, we propose an effective greedy heuristic (GH) including two phases (task ordering and task scheduling). GH uses an Earlier Latest Start Time First method (ELSTF) for task ordering with the result that a task sequence is obtained. A Task Dispatching method (TD) is established for the task scheduling, in which each task in the obtained task sequence is scheduled one by one. Experimental results demonstrate that the proposed GH outperforms the baseline (RoundRobin) remarkably. ELSTF and TD are also verified to be effective.