Arrebol - sistema elástico de processamento em lote de aplicações single-node.

Abstract

With the advancement of computers and their increasing use in many different areas, research that uses computing intensively to obtain results, also known as e-researches, have become increasingly present in academia and in companies that need to run applications that process large amounts of data. For a long time, the most common environment for running these applications was the supercomputers, which are infrastructures with high installation and maintenance costs and are mostly present in specific locations, such as large companies and academic institutions. Therefore, this type of infrastructure is not always accessible to all researchers. Supercomputers have previously allocated computational resources (CPU, RAM, storage) that if not properly used, either because of an inefficient scheduling policy or low demand, end up being wasted. With this in mind, we propose Arrebol: a cloud-based batch processing system that takes advantage of the elasticity of this environment and allocates resources on demand. Arrebol monitors the cluster's workload and tries to balance it with the available resources. Whenever possible, it tries to optimize resource utilization by keeping it as small as possible. The present work continues the Arrebol development by implementing the resource management server (Resource Manager).