ECOS: An efficient task-clustering based cost-effective aware scheduling algorithm for scientific workflows execution on heterogeneous cloud systems

Cloud Computing provides an attractive execution environment for scientific workflow execution. However, due to the increasingly high charge cost of using cloud service, cost minimization for workflows execution on cloud systems has become a crucial issue. Traditional work are adopting the sophisticated scheduling techniques to address such issue. Differently, this paper has proposed an efficient task-clustering based cost-effective aware scheduling algorithm (ECOS) to minimize the cost without comprising the deadline constraint. First, with respect to the characteristics of multi-type workflows, cloud heterogeneity and cost model, we have formulated the problem of task-clustering to simplify the structure of workflows and workflow scheduling to minimize cost within the deadline constraint. Then, we have devised ECOS with two key steps: (1) vertical clustering is with the time consideration that selectively merges the sequential tasks to reduce the transferring time within the workflow; (2) horizontal clustering and greedy allocation is to aggregate the parallel tasks and greedily allocate resources to that tasks with the aim of minimizing cost within deadline. Last, we have conducted the experiment that compare with well-known task-clustering based algorithms via Workflow Sim platform. The results have demonstrated that ECOS can efficiently merge tasks and minimize the total cost without comprising the deadline constraint both in small and large datasets. Moreover, we have discussed the ECOS in terms of various schedulers and number of tasks to validate the performance of ECOS. (C) 2019 Elsevier Inc. All rights reserved.