Abstract:

In scientific workflows, cluster or grid platform is used to deal with complex problems, and the emergence of cloud computing offers an alternative. The transferred data are huge in the dataintensive scientific workflows in cloud computing. Reducing the count and amount of the data transfers between different clusters in a datacenter is a challenging problem. There exists dependency in data of the scientific workflow. Firstly, a relational matrix is built based on the associate amount. Secondly, the relational matrix is clustered using bond energy algorithm. Thirdly, the clustered matrix is partitioned to k parts and each part is deployed to corresponding clusters in the datacenter. The simulation results show that the proposed method reduces the count and amount of the data transfers between clusters in a datacenter.

Key words: cloud computing;associated amount;bond energy algorithm;data placement;task scheduling