Creating science-driven computer architecture: A new patch to scientific leadership Page: 2 of 6
This report is part of the collection entitled: Office of Scientific & Technical Information Technical Reports and was provided to UNT Digital Library by the UNT Libraries Government Documents Department.
Extracted Text
The following text was automatically extracted from the image on this page using optical character recognition software:
We recognize that the computer industry cannot participate in the long term unless sufficient
markets exist for these machines. Therefore, while our central goal is faster scientific application
time-to-solution, our joint goal is to develop architectures that industry can adapt to other markets
to assure the sustainability of this national effort for science.
In this white paper we propose a strategy for accomplishing this mission, pursuing different
directions of hardware development and deployment, and establishing a highly capable
networking and grid infrastructure connecting these platforms to the broad research community.
2. A Strategy for Creating a New Class of Computer
Architectures for Scientific Computing
In the 1980s and early 1990s there were more than twenty U.S. companies producing
supercomputers that were designed for scientific and technical applications. Among them were
Alliant, BBN, Convex, Cray Research, Cray Computer, IBM, Intel Supercomputer Division,
Kendall Square, and Thinking Machines. The primary role of national laboratories and
universities at that time was to establish performance requirements, evaluate offerings and select
the best for use in scientific research. The market for high-performance scientific and technical
computing was a significant focus of the computer industry.
Today the situation is radically different because the market for commercial Web and data servers
has grown to completely overshadow the market for high performance scientific computing.
Supercomputers in use now largely consist of clusters of commodity servers, connected by
networks that have not increased in capability at the same rate as the processors they connect.
Today's situation calls for a strategy that creates a new class of supercomputing machinery by
leveraging Moore's Law and the technology that underlies commercially viable computers and
the microprocessors they are based on, instead of simply utilizing its existing implementations.
Most crucially, this strategy must provide a new way to couple scientific applications
requirements to the development of computer architectures, thereby opening a sustainable path to
petaflops/s-level performance and beyond.
2.1 Sustained Cooperative Development of New Computer
Architectures
We propose a new type of development partnership with computer vendors that goes beyond the
evaluation of the offerings that those vendors are currently planning for the next decade. We
therefore propose in this paper a comprehensive strategy that includes development partnerships
with multiple vendors. Those partnerships will bring to bear:
1. teams of scientists and computational mathematicians who will modify and optimize their
applications for future systems through the use of performance modeling, simulators and
prototypes of new hardware
2. teams of computer architects from major U.S. computer vendors who will interact directly
with the scientific applications teams, and
3. teams of computer scientists who will work with both applications scientists and computer
architects to analyze and abstract the requirements of scientific applications so that they can
be addressed in hardware and to develop the software environments that will allow scientists
to extract the maximum performance and capability from that hardware.
This strategy is directed at challenging and partnering with vendors to create architectures that
perform to a target level on a specific suite of scientific applications. Unlike the current approach
Upcoming Pages
Here’s what’s next.
Search Inside
This report can be searched. Note: Results may vary based on the legibility of text within the document.
Tools / Downloads
Get a copy of this page or view the extracted text.
Citing and Sharing
Basic information for referencing this web page. We also provide extended guidance on usage rights, references, copying or embedding.
Reference the current page of this Report.
Simon, Horst D.; McCurdy, C. William; Kramer, T. C.; Stevens, Rick; McCoy, Mike; Seager, Mark et al. Creating science-driven computer architecture: A new patch to scientific leadership, report, May 16, 2003; Berkeley, California. (https://digital.library.unt.edu/ark:/67531/metadc735094/m1/2/: accessed April 17, 2024), University of North Texas Libraries, UNT Digital Library, https://digital.library.unt.edu; crediting UNT Libraries Government Documents Department.