Research On Cloud Workflow Task Scheduling Method Based On Heuristic Algorithm

At present, most of the cloud workflow scheduling algorithms mainly consider the individual requirement of users. Commonly using heuristic algorithm to solve the optimization problem of a single target. In the user view, execution time and execution cost are the most important performance indicators. They hope the application that they submit could be completed successfully before the deadline and the total execution cost is minimal. In addition, most of the current scheduling algorithms that are based on time optimization ignore the resource reliability, which leads the failure of task, thus causing extra loss of time. In view of the above questions, this paper focuses on how to minimize execution cost within deadline and how to choose the resource with high reliability for the task to ensure a high success rate of task execution.According to the relationship between execution time and execution cost of the task, we propose the deadline constraint-based scheduling algorithm for cost optimization. It firstly assigns the deadline to the latest finish time, the earliest start time and the earliest finish time of each task. Then obtaining the critical parent task and partial critical path of the end task. Allocating a resource to each task of the partial critical path when it could finish the task before the latest finish time and the cost is minimal. During the execution, it adjusts the time constraints of other tasks dynamically until the completion of resource allocation of all tasks. The algorithm achieves the cost optimization within deadline.Aiming at the relationship between execution time and the resource reliability, we propose the scheduling algorithm for cost optimization based on reliability. The algorithm introduces the reliability model including resources reliability and transmission reliability. It treats the reliability and execution time as the optimization goal. It firstly produces the Pareto front set of multi-objective optimization. Then selecting the best scheme according to the preference of user form Pareto front set as the final scheduling scheme. The experiments shows that the algorithm not only reduces the search space and search time, but also decreases the total execution time while the number of task is very large or the network is congested. Thus it achieves time optimization.