Home // International Journal On Advances in Systems and Measurements, volume 17, numbers 1 and 2, 2024 // View article

Optimized Hardware Procurement for High Performance Computing Systems

Authors:
Scott Hutchison
Daniel Andresen
William Hsu
Mitchell Neilsen
Benjamin Parsons

Keywords: HPC; Procurement Optimization; Recommender system; XGBoost.

Abstract:
When faced with upgrading or replacing High Performance Computing or High Throughput Computing systems, system administrators can be overwhelmed by hardware options. Servers come with various configurations of memory, processors, and hardware accelerators, like graphics cards. Differing server capabilities greatly affect their performance and their resulting cost. For a fixed budget, it is often difficult to determine what server package composition will maximize the performance of these systems once they are purchased and installed. This research uses simulation to evaluate the performance of different server packages on a set of jobs, and then trains a machine learning model to predict the performance of un-simulated server package compositions. In addition to being orders of magnitude faster than conducting simulations, this model is used to power a recommender system that provides a precision@50 of 92%. This model is further evaluated using 24 days throughout the calendar year, and it achieves a precision@50 of 88%.

Pages: 46 to 55

Copyright: Copyright (c) to authors, 2024. Used with permission.

Publication date: June 30, 2024

Published in: journal

ISSN: 1942-261x