| This paper investigates a workflow scheduling problem with stochastic workflow arrival times,uncertainty task processing times,uncertainty data transmission times and uncertainty due dates.The problem is common in many real-time and workflow-based applications,where tasks with complex DAG structureare executed on scalable cloud resources with multiple price options.In the existing works,most of them assum that the information considered in clouds is assumed to be static and can be estimated in advance.However,there are many uncertainties in the real cloud environment,such as uncertain task start/end time,the new workflows arrive time and deadline due to performance fluctuation of virtual machines and the uncertain scheduling environment.If these uncertainties are not taken into account,it may lead to a workflow cannot finish before a due date or renting more cloud resources.An adaptive iterative heuristic workflow scheduling method is proposed in this paper.The -quantile based model is formulated.The main optimization goal is to minimize the cost.This method includes three different scheduling stages.Firstly,due to the random arrival of scientific workflows,workflow scheduling algorithm should be able to collect newly arriving tasks.And then,this paper proposes three priority strategies based on the uncertainty temporal parameters of tasks including the earliest start time,earliest end time and Virtual Machine(VM)resources.A re-schedule strategy is applied to improve the schedules.Then an Uncertainty Dynamic Event Scheduling,named UDES,to periodically schedule tasks.This paper defines Optimistic Task Deadline(OTD)and Pessimistic Task Deadline(PTD)in order to achieve a better effect.Lastly,experimental results show that the proposed algorithm is more effective and robust than the existing methods. |
