Sistema modular de adquisición, gestión y visualización web de métricas de desempeño para centros de cómputo de alto rendimiento. Caso de estudio CECAD

Abstract

The purpose of this project is the design, implementation and evaluation of a modular system of acquisition, management and web visualization of performance metrics for high performance computing centers (HPC). Metrics such as: CPU usage, RAM memory utilization, disk space, process counting, network traffic, disk I / O operations, number of threads, memory usage in GPU are collected through collectd, metricbeat and statsd which are grouped by collectd and sent to a persistence layer supported by Elastic search and finally visualized using Kibana. The system will have a modular design in order to adapt to the different scenarios of high-performance computing centers and its continuous growth and inclusion of new technologies. The implementation will be a case study specifically for CECAD.