A Randomization Approach for Stochastic Workflow Scheduling in Clouds

Algorithm 1

The outline of the randomization scheduling approach.

Input: DAG on a set of resources with stochastic model

Output: A schedule specifying task mapping and execution order

 (1)  Create a candidate schedule list , which is initially empty.

 (2)  Run a deterministic heuristic DH to generate the schedule entirely based on all mean values of task execution

     times and communication times, and put the schedule into .

 (3)   while the termination condition of the producing phase is not met repeat

 (4)    Run the randomized heuristic RH and push the result schedule into .

 (5)   endwhile

 (6)  Compute the expected makespan based on mean values of all stochastic inputs for each schedule in .

 (7)   Return the schedule with the minimum expected makespan as the result schedule.