NHSE ReviewTM 1997 Volume First Issue

Establishing Standards for HPC System Software and Tools

| <- HREF="abstr.html" Prev | Index | Next -> |
NHSE ReviewTM: Comments · Archive · Search

Chapter 1 -- Introduction

The number and variety of parallel computing systems has grown dramatically over the last decade. The good news is that as more systems have been deployed in research and industry settings, our collective experience base in parallel programming has reached a respectable level. Today's application developers have access to a broader range of support - in terms of established parallel algorithms, programming languages, and system software - than was the case even a few years ago. That support is also considerably more robust than it was in the early days of parallel computing [cf. 6, 19].

Unfortunately, the quality of system software and tools on high- performance computing (HPC) machines remains far below the expectations and needs of the user community. The most visible consequence is that HPC has not been a high-growth market; a number of manufacturers have been driven to bankruptcy or to mergers with companies capable of drawing on larger consumer bases. Less visible, but of no less impact on the national economy, is the fact that unacceptably high levels of time and effort are required to develop HPC applications [15].

Indeed, too much effort is required just to keep an existing HPC application viable beyond a year or two. A report from the National Science Foundation began its discussion of barriers to the future progress of HPC with this statement:

The need for a robust, consistent software infrastructure across HPC vendors and platforms has been a recurring theme at many workshops and conferences over the past few years. In 1995, a national task force was convened to establish the basic requirements for such an infrastructure. As part of the task force effort, we prepared a Web-based document entitled "Guidelines for Writing System Software and Tools Requirements for Parallel and Clustered Computers". The report was also released as technical reports by JPL and Oregon State University [18]. It lays out standard verbiage for system software and tools requirements, and is intended for use in preparing RFPs and other procurement documents. Because the intent of the document is to serve as a blueprint for writing procurements, it is a formally worded checklist. No attempt was made to explain why particular software is needed, or how the group arrived at particular decisions.

This article provides that background information. It outlines the general context of the task force in terms of the importance of software standards to the future of HPC. Chapter 2 addresses the methods that have proven successful for establishing software standards in the past. The next chapter discusses the somewhat unique situation presented by HPC software, giving rise to the somewhat unorthodox approach taken by the task force.

In Chapters 4 through 9, six major categories of system software and tool support are outlined. Each chapter describes the group's rationale for establishing guidelines, including where software elements were omitted because the technology is not mature enough to permit standardization or because the need did not appear to span a full range of user sites. In addition, an attempt has been made to capture the group's discussions regarding both obstacles to standardization and particularly fruitful areas for further work.

Copyright © 1996

| <- HREF="abstr.html" Prev | Index | Next -> |
NHSE ReviewTM: Comments · Archive · Search

Copyright © 1997 Cherri M. Pancake