一种科学工作流视图抽象方法

View abstraction method of scientific workflow

为了解决复杂的科学工作流过程视图造成用户理解和数据溯源困难的问题,在对科学工作流及其视图进行形式化定义的基础上,提出一个可保持数据溯源正确性的视图抽象算法——Extend-and-Merge算法。该算法从科学工作流中的一个任务出发,通过寻找及合并三种合理子结构,最终得到保证溯源正确性的合理抽象视图,并设计实现了一个软件工具用于可视化展示。通过一组真实科学工作流过程数据上的实验表明,Extend-and-Merge算法可在多项式时间内获得规模为原来过程13.1%的合理抽象视图。

It is difficult for users to understand and analysis the provenance of data with complicated and enormous scientific workflow. To solve this problem, a view abstraction algorithm of Extend-and-Merge was proposed to maintain provenance correctness based on the formal definition of scientific workflow and workflow view. In this al- gorithm, three kinds of sound substructure were searched and merged from one task of scientific workflow, and the reasonable abstract view to ensure the provenance correctness was obtained. Meanwhile, a software tool was de- signed to realize the visual presentatiorL Experiments were performed on a data set of real scientific workflow process, and the result showed that the proposed algorithm could obtain the reasonable abstract view to original process 13.1% in polynomial time.