A new data placement approach for scientific workflows in cloud computing environments

Abstract

The reach of Cloud Computing technologies approved distributing with massive data applications such as Scientific Workflows, which processing huge scientific data in dispersed computing infrastructures. Among the characteristics of Cloud Computing, we mention the elasticity that allows workflows to dynamically stipulate necessary resources for tasks execution. The processing of massive data with scientific workflows increase the data transmission, rise execution delay and it request huge bandwidth cost. So, to reduce the execution cost of workflows and the data movements, data placement optimization technics must be taken into consideration. While placing datasets during execution of tasks for a job in a workflow, there are dependencies between datasets and between tasks. In this paper, we propose a data placement approach based on heuristic genetic algorithm which takes into accounts control and data flow dependency, in order to reduce data movements and so the utilization of resources in cloud environments.