Workflow scheduling method based on critical path task prospects

Abstract

The invention relates to a workflow scheduling method based on critical path prospects. The method comprises the steps that a user submits workflow; the workflow is converted into a directed acyclic task model graph (DAG); and scheduling of DAG task nodes is performed, and a workflow scheduling scheme is output. According to the method, a longest path from an entrance task to a current task and a longest path from the current task to an exit task are considered at a task priority determination stage; and influences of critical path tasks and non-critical path tasks on a scheduling result are considered at a resource selection stage. Compared with other methods, completion time of workflow scheduling is short through the method.

Description

technical field [0001] The invention relates to the technical field of workflow scheduling in a cloud computing environment, and mainly relates to a workflow scheduling method that comprehensively considers the task itself and its subtasks when allocating resources for critical path tasks. Background technique [0002] In recent years, a variety of loosely coupled heterogeneous distributed computing models or new technologies have emerged, such as cloud computing. With the rapid rise of cloud computing, more and more workflow applications are based on cloud computing platforms to improve computing speed. Each workflow contains a task set, how to allocate tasks in each task set to appropriate heterogeneous computing resources more efficiently and quickly is the research focus of workflow scheduling problems. The static workflow scheduling problem generally expresses the task relationship in each task set using a DAG (Directed Acyclic Graph) graph before scheduling, and then ...

AI Technical Summary

Problems solved by technology

Although the scheduling results given by the HEFT algorithm are generally ideal, the algorithm only considers the impact of the tasks to be scheduled and their predecessors; the main idea of ​​the CPOP algorithm is to find the critical path of the DAG graph, and then define a critical path processor. This processor needs to meet the minimum calculation time of all tasks on the critical path, and finally all the critical path tasks are scheduled to this critical path processor in the scheduling, but the performance of the CPOP algorithm is not good

In 2010, L.F.Bittencount, R.Sakellariou and E.R.M. Madeira proposed the Lookahead algorithm, which is an improved algorithm for the HEFT algorithm. It mainly considers the impact of task resource allocation decisions on its subtask scheduling during the scheduling process. Experiments It shows that the performance of the Lookahead algorithm has been improved, but it also increases the time complexity

In 2014, H.Arabnejad and J.G. Barbosa improved the lookahead algorithm for the long execution time and proposed the PEFT algorithm. The main idea of ​​the algorithm is to construct an optimistic time-consuming table before scheduling, and predict the subtasks of tasks when they are scheduled on each resource. Scheduling results to optimize resource allocation decisions. Although the time complexity of the PEFT algorithm is reduced, the scheduling effect is not good

Images