Computing Reviews

HPC forecast : cloudy and uncertain
Reed D., Gannon D., Dongarra J. Communications of the ACM6682-90,2023.Type:Article
Date Reviewed: 05/08/23

This article points to possible directions for high-performance computing (HPC) ecosystem development and the related big challenges for the scientific community and the future of advanced scientific computing. The authors note that, over the last 20 years, this direction was governed by advances in semiconductors, but now we are more and more dependent on hyperscale cloud vendors who are creating new software and hardware technologies.

For the communities of scientists and engineers, Reed at al. provide a detailed analysis of the current state of HPC and possible directions for its development, taking as a reference point what is seen in other areas of computer engineering. Since we are entering a new era of cloud services it should be expected that this significant paradigm shift will also be visible in the case of HPC. As an additional argument, the authors give examples of large-scale scientific instruments (the Vera Rubin Observatory, the Large Hadron Collider, or the LIGO gravity wave detector) that were built to explore different scientific questions by the collaborating community. This was possible thanks to innovative government and private sector partnerships. The same should be expected for HPC centers, for example, sharing common costs. While the authors’ view can be disputed, it is supported by data and strong arguments and gives five maxims as good guidelines for government and private sector research and development (R&D) strategies.

As usual, it is a matter of money: we can continue to develop and deploy new generations of HPC systems, but there is a need to invest hundreds of billions of dollars. In the authors’ opinion, the answer to this challenge is public-private partnerships that find a balance between commercial goals and scientific expectations. Within the next ten years, we can check whether or not the authors’ expectations are confirmed. However, no matter what results we can expect in the future, this article is a good and valuable summary of the current state. It provides both the public and private sectors with an in-depth reflection on mid- and long-term strategies related to broadly understood digitalization.

Reviewer:  Dominik Strzalka Review #: CR147587 (2307-0095)

Reproduction in whole or in part without permission is prohibited.   Copyright 2024 ComputingReviews.com™
Terms of Use
| Privacy Policy