Cloud workflow task clustering method for supporting dependency balance and time balance

Abstract

The invention discloses a cloud workflow task clustering method for supporting dependency balance and time balance. According to the method, when tasks of the same level in a flowchart are clustered, data dependency relationships among the tasks are considered at first, the time balance of clustering is considered, when the tasks of the same level are clustered, parent tasks with the same sub-tasks are not simply gathered together, the common sub-tasks of the parent tasks and respective special sub-tasks of the parent tasks are comprehensively considered, and on this basis, a concept and calculation formula of the degree of task correlation are put forward to characterize the dependency degree among the tasks; meanwhile, on the basis of considering the dependency relationships among the tasks at first, the running time of each task is further considered, and the dependency balance and time balance of clustering are ensured simultaneously. Through experimential comparison with traditional clustering methods, it can be found that clustering the tasks by the adoption of the cloud workflow task clustering method for supporting the dependency balance and the time balance can more effectively shorten the completion time of workflows.

Description

Technical field [0001] The invention belongs to a workflow running time optimization technology in cloud workflow scheduling. When clustering tasks at the same level, based on the data dependency between tasks and the time balance between clusters, a cloud workflow task clustering method that supports dependency and time balance is proposed. Background technique [0002] Research in various scientific fields (such as physics, astronomy, and biomedicine) requires running large-scale applications to process data. These large-scale applications are usually workflow applications. These workflow applications are composed of many computationally intensive tasks with dependencies, and the amount of calculation of these tasks varies greatly. Some tasks only take a few seconds to run, while others take several hours. When these short-running tasks are executed in a cloud computing environment, there will be significant scheduling overhead, and the extra time generated by the scheduling o...

AI Technical Summary

Problems solved by technology

This timing imbalance will delay the start time of the next level task

Images