抽象网格工作流中数据集划分方法的研究

Research of Datasets Division in Abstract Grid Workflow

通过分析AGWL中已有的数据集划分方法,发现在涉及到数据子集相关性时,现有的研究方法仅是通过一种定性的复制数据块间部分子数据集的方法进行处理的,并没有基于数据子集相关性给出合适的划分方法。针对这一状况,提出了一种使用矩阵对数据集进行划分的方法。首先用矩阵的下标表示子数据集,矩阵值表征相对应的子数据集的相关性,然后通过对矩阵值与一个设定的阈值的比较,对矩阵进行归并处理,最终实现了一种基于数据子集相关性和灵活性的数据集划分方法。

Analyzing the existing methods for the datasets division in abstract grid workflow （AGWL）shows that relating to correlation of data subsets,the existing methods only qualitatively replicate some data subsets between data blocks for processing,and no appropriate division method can be found based on the correlation between the data subsets,in the view of that,making use of matrix to divide datasets was proposed.The matrix index can indicate sub datasets,and the matrix value represents the correlation between the data subsets,and comparing matrix value with a given threshold can process and merge the matrix so that a flexible method for datasets division can be reached based on the correlation between the data subsets.