Monitoramento de recursos computacionais em Cluster de alto desempenho: implementação e avalia ção com foco na sustentabilidade

Abstract

Implementing a high-performance cluster computing monitoring system is a beneficial tool for the strategic management of computing resources, given that monitoring provides an understanding of the behavior of resources, enabling various analyses. In this sense, considering that high-performance computing systems are scarce equipment, with few centers available due to the high cost, maintenance, energy and cooling; strategic management to improve and increase their lifespan is necessary. Furthermore, observing the context of the Legal Amazon where this work was developed, where issues such as sustainability and reduction of carbon dioxide emissions are crucial, the monitoring of computing systems becomes an extremely important tool for adopting strategies that aim to reduce the carbon footprint in the atmosphere. The literature presents several monitoring tools, however, there is a gap between systems that use OpenPBS resource management about the collection of metrics through Prometheus, which made it necessary to develop a customized exporter in line with OpenPBS resource manager monitoring commands, displaying customizable dashboard graphs through the Grafana data visualization tool, which found 20 different surveys related to node availability, resources consumed by users and employment status. This work aims to contribute to the sustainable development of Amazon and implement a computational resource monitoring system as a precursor tool for the introduction of green computing in the cluster aligned with the Institutional Development Plan. In this perception, the data of interest from the OpenPBS resource manager were identified, obtained, stored and made available for visualization in graphs. Therefore, this work is relevant as a product for specific administrators who want to understand the behavior of the cluster.