A Walking Dwarf on the Clouds
Laion F. Manfroi, Mariza Ferro, André Yokoyama, Antonio Roberto Mury, Bruno Richard Schulze · 2013
Scientific computing often requires high performance and distributed computational resources to perform large scale experiments in order to achieve accurate results in due course. These needs have been addressed with dedicated high performance computing (HPC) infrastructures. New models to achieve high performance are the combined multi-core architectures with accelerators (manycore). A new dimension is also added by the Cloud Computing model, which has garnered significant attention from industrial and scientific community, as a potential model to address a broad array of computing needs. Even so, it is necessary to evaluate how scientific applications behave on these models. In this work, we propose an approach based on the Dwarfs classification to evaluate architectures, considering that applications have different characteristics in terms of resource consumption. The results show how the Dwarfs approach could help in choosing the architecture and in Cloud Computing environment evaluation when looking for better performance.