Neblina - Scientific management environments for clouds

Abstract

Scientific computing often requires high performance computational resources to perform large scale simulations in order to achieve appropriate results. These demand have been addressed with dedicated High-Performance Parallel and Distributed Computing (HPDC) infrastructures, but nowadays, a new dimension has been added by Cloud model, which has gathered significant attention from scientific communities, as a potential approach to address a broad range of computing demands. The challenge now is how to provide the correct balance between real and virtualized infrastructure, able to comply scientific application requirements. Our Neblina approach presents a model combining real and virtual environments, with scientific applications being scheduled on the resources based on the application categorization, the virtualization layer capabilities and the hardware infrastructure characteristics, aiming to optimize the use of the computational resources, minimizing the loss of performance, due the virtualization layer.