Data Placement Based On Hierarchical Clustering on Scientific Workflows

Data play the main role in scientific workflows. In the cloud environment there are many workflows need these data and their size might be exceeded to terabytes or petabytes. Since these workflows consist of many interdependent tasks and each task in the workflow requires some dataset as its input, the data should be somehow managed in order to produce decent results in both task execution and data movements. The required datasets might be placed on different locations, hence, the required datasets for a task needs to be retrieved and positioned in the destination host. It causes data movements and makes some delay on the task execution. In these paper we study a kind of clustering, called hierarchical, and used it as an approach for better data placement. The performance of this method is compared with random data placement and an extended genetic algorithm. The results show about 20% improvement is obtained against random data placement.

کلیدواژه ها:

Keywords—Hierarchical Clustering ، Data Placement ، Scientific Workflows ، Data Management on Cloud Environment