PDCO 2022 Keynote Talk: Performance and Energy models for modern HPC servers.

IEEE International Parallel and Distributed Processing Symposium (IPDPS)(2022)

Cited 0|Views8
No score
Abstract
Most energy-aware studies are based on hypothesis. The most important are the models of the studied system. Depending on the precision of these models, the proposed theoretical framework can be transposed on actual system or are limited to theoretical studies. During this keynote, we will explore the differences between classically used models of performance, power, and energy of HPC systems on one side, and actual measures on several clusters on the other. It will include elements on the heterogeneity of homogeneous systems. We will also propose more realistic models for expressing the behavior of HPC systems, from hardware to applications. Some elements on the tradeoff between the complexity of the algorithms and their precision will also be precised. One of the key metric of this complexity will be the amount of data needed to obtain them as measuring energy-related information on HPC datacenters is often difficult.
More
Translated text
Key words
modern HPC servers,energy-aware studies,HPC systems,actual measures,homogeneous systems,energy-related information,HPC datacenters,energy models
AI Read Science
Must-Reading Tree
Example
Generate MRT to find the research sequence of this paper
Chat Paper
Summary is being generated by the instructions you defined