Info: Zenodo’s user support line is staffed on regular business days between Dec 23 and Jan 5. Response times may be slightly longer than normal.

Published February 1, 2018 | Version v1
Technical note Open

Ferramentas de Análise de Desempenho e Energia: Visão geral, instalação e uso

Description

Scientific computing generally requires huge processing power resources’ to perform large scale experiments and simulations in a reasonable time. These demands have been addressed by High Performance Computing (HPC), which makes them invaluable tool for modern scientific research. Nowadays, the simulations performed on the petascale supercomputers can achieve accuracy and fidelity never seen before. Despite this, several areas require more computational power to perform their simulations. But, achieving exaflop capabilities imposes some challenging tasks, such as the performance evaluation in HPC, the understanding of the computational requirements of scientific applications, its relation with power consumption, and ways to monitor a wide range of parameters in heterogeneous environments.

       To this end, we were finding solutions to enable the monitoring of these relevant parameters for analysis of scientific applications in HPC environments. This is not a trivial task, because it imposes to measure a set of distinct parameters and there isn’t an unique monitoring tool able to capture all these aspects and provide them for analysis. For this reason we investigated a range of monitoring tools to enable a high-resolution profile of performance and power consumption with low overhead.

       Throughout this study it was noted other difficulties, which can greatly increase the time to the development of research, such as: i) each tool has different approches to measure performance (online, offline, sampling, timing, counting and tracking) and power; ii) different dependencies and complexities for the installation process; iii) works on only for some operational system or progranmm language and iv) collect a different set of parameters;

       Therefore, this report presents different performance and power monitoring tools, addressing, for each of them, the techniques used for data collection, the structure of the tool for its operation and how the data obtained are presented to the user for analysis. In addition, the details for the installation and the use of the tools are also presented. The main contribution of this report is to be a tutorial for how to install and use, as this process is often costly, since most tools have many dependencies and configurations. Despite these, we present our experience when using the tools and for some of them pointing out its strengths and limitations.

Files

Ferramentas de Análise de Desempenho e Energia: Visão geral, instalação e uso.pdf