Dated reports of innovation are inevitably eclipsed by other, newer advances that tend to play out rapidly in either the scientific or popular press. Thus, this report is also dated in the scope and freshness of announced results. Nevertheless, these presentations do offer a glimpse into certain trends that underscores the acceleration of scientific enterprise. In general, the pace of innovation appears to be quickening, with multiple advances in multiple fields leading to yet greater technological acceleration. Nothing is ever certain, but if the rate of applied innovation continues to increase, even the most optimistic forecasts may prove in retrospect hesitating and timid. Although rearranged thematically and edited for clarity, the sections that follow contain material from the original papers published by the cited authors.


As computational power continues a seemingly inexorable advance, interest in and exploitation of new microprocessor architectures and software techniques remain strong. Computing is increasingly Internet-centric, with computer nodes distributed among desktops rather than in climate-controlled repositories. As Gharachorloo notes, administrators are finding it easy and inexpensive to scale up computational power by installing multiple independent systems. Although management issues inevitably arise, making more systems available simultaneously in a loosely clustered environment allows for incremental capacity expansion. The accelerating growth of the World Wide Web should continue to encourage the development and deployment of high-performance computers with high availability and incremental scalability. Hardware improvements are, however, proving easier to implement than is the software to support those improvements.1

Outside the purview of the FOS/FOE symposia are advances in biological and optical computing, both of which must be said to be in the very earliest stages of development. Nevertheless, their potential, either as stand-alone systems or integrated in some as-yet-unanticipated fashion, could have a substantial impact on future computer design and deployment. Even though robust real-world architectures for either have yet to be perfected, their promise, in terms of sheer computational power, is orders of magnitude beyond current serial-processing applications, and they should be considered as future contributors to advanced computing initiatives.

More promising still is quantum computation, which employs individual atoms, molecules, or photons in exploitation of quantum interference to solve oth-


Frontiers of Engineering/1999. “Evolution of Large Multiprocessor Servers,” Kourosh Gharachorloo, pp. 11-19. Many of the FOE papers cited in this report can be found at <>.

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement