Abstract
Intensive workflows require a lot of computational resources and massive data movement between Storage and Computing servers hosting these datasets and tasks. Moving these datasets among these servers may increase the execution time, generate a high energy consumption by communication devices and a significant data movement cost. Thus, we need a good data placement strategy (DPS) to minimize the data movement between these servers, the communication energy consumption and the workflow execution time and cost. In this paper, we propose a data placement strategy based on the Formal Concept Analysis approach that considers the original datasets, the different communication levels (switches, routers) and the granularity of used resources in data center. It aims at grouping the maximum of datasets and tasks in a minimal number of Storage computing Servers (SC) as close as possible to each other. Simulations show that our strategy can greatly reduce the energy consumed by communication devices, the volume of data movement as well as the execution time and cost of the workflow.