A resilient and cloud-based batch processing system for high performance computing.

Abstract

Since computers were able to execute more than one program at a time, the batch systems became usual. In this context, many batch applications require a high level of processing capacity, which leads us to high performance computing. This approach has been used for long, mainly for scientific purposes. It is common that the conventional environments for HPC, which are local clusters and supercomputers, provide a command line interface for the users to enable them to run their applications in nodes connected through high-speed networks. However, high-speed networks might not be useful in single-node applications, becoming a waste point. Besides, despite the old users being used to the command line, it may frustrate newcomers. That's why we propose, with this work, a HPC supported system that takes advantage of some cloud's features to minimize the cost and the waiting time, while focusing on the user experience.