Profit-Oriented Scheduling Optimization for Workflow in Clouds

Adaptive workflow scheduling for diverse objectives in cloud environments

“…With the method in , L F T for each task and CP can be computed. We set the rules as : Task with only one available resource should be assigned first. Finish the task before its LFT. Task with bigger out‐degree has higher priority. Task directs to a task with bigger in‐degree has higher priority.In DRAWS, all the tasks, which are schedulable, have a priority by rules 1–4. The former rule has higher priority.…”

“…With the method in [16], LFT for each task and CP can be computed. We set the rules as [25]: Rule 3: Task with bigger out-degree has higher priority. Rule 4: Task directs to a task with bigger in-degree has higher priority.…”

“…We generate the task list L T . The tasks in L T are all schedulable tasks and they are sorted in descending order by priority (lines [16][17][18][19][20][21][22][23][24][25][26][27].…”

“…If needed, choose one with best performance to start a VM based on the task demands and add the scheme into the solution space. Otherwise, reject the task (lines [11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28]. And in the algorithm, we set the VM on a busy host with higher priority (line 34).…”

“…In this section, we talked about the solutions of multi‐objective scheduling problems and found out that best scheme choosing under different conditions is the most challenging problem. Based on former work , we find that converting the workflow into a task list with priority, which is decided by the topologic characteristic, is an effective method to solve the precedent constraints challenges. For the well‐known no free lunch theorem , it is impossible to find an optimum algorithm under multi‐objectives without the prior structural knowledge.…”