Computing Facilities
The Department contains a large
array of sophisticated computer hardware which supports both our
research and instructional missions. This equipment is maintained by
a central facility, the Computer Systems
Laboratory. We are continuously upgrading and enhancing our
systems to offer the most up-to-date computing resources possible.
Much of the equipment was donated by our industrial affiliates; their
support has been invaluable in enabling us to develop a first-rate
computing environment.
All faculty, supported graduate students (TAs, RAs, and Fellows), and staff
have high-performance workstations on their desks. These include various
models of DEC, SUN, HP and Intel-based PC's. Desktop workstations
run various flavors of the Unix operating system or Windows NT.
Parallel and Distributed Computing Facilities
The Department is recognized as a national leader in research on
parallel and distributed computing. Current work involves experimental
design of both parallel algorithms and computer architectures in
support of a wide range of projects, including mathematical
programming, parallel-program debugging tools, performance modeling
and analysis, computer vision, databases and many others. The TOPAZ,
PRISM, and MIDSHIP projects, funded by NSF Institutional Infrastructure grants,
have enabled us to acquire parallel hardware to enhance
this work. Components of our parallel computing environment include a
12-processor Sparc Enterprise 5000, two 8-processor Sparc Enterprise
5000's, four 16-processor Sparc Enterprise 6000's, and the Cluster of
Workstations (COW),
The Wisconsin COW
which is a collection of forty Sun SPARCstation-20 servers connected
by a high-speed network. Each workstation has two processors and
enough memory (64 MB) and local disk space (2 GB) to run large
programs. COW supports a wide range of experiments and is itself an
experiment in building supercomputers using the highest performance
computing and networking components available, with little custom
hardware.
A recent addition to this parallel infrastructure is the Cluster of
Personal Computers (COPcS), which is a collection of 20 dual-processor
Pentium's connected with fast ethernet. Each PC has 128 MB of memory
and 12 GB of local disk space for large database experiments.
CONDOR Distributed Resource Management
A locally developed software package called
Condor provides additional
computing power for compute-bound tasks such as simulations. Condor
automatically locates workstations which are idle and transfers jobs to them.
The jobs are periodically checkpointed and migrate from machine to machine
until completion. Studies of Condor showed that jobs
submitted to it made use of over 180 CPU-days per week of otherwise wasted
machine cycles.
Network Services
Most of our research and instructional facilities are connected to local area
networks, each of which is connected to every other and to the Internet by
routers. The network allows remote and automated use of departmental
resources and information sharing. There is currently over 800 Gigabytes of
storage available to most of our machines though the
AFS distributed file
system. Much of the research information produced by the Department is made
freely available to the world through
our World-Wide Web server.
Instructional Facilities
In addition to the research facilities, the Department has a number of
workstations to support work in undergraduate and graduate courses. These
include forty UltraSPARC-10 and forty SPARCstation-20 workstations.
During the past two years, Hewlett-Packard has donated a laboratory of
200-Mhz Pentium Pro workstations and Intel has donated a laboratory of 400 Mhz
dual-processor Pentium II workstations to support undergraduate computer
sciences education. Additional instructional equipment, used for introductory
programming courses, include sixty Pentium Pro computers. Our instructional
laboratories support more than 2000 students each semester.
The Department also supports an
Undergraduate Projects Laboratory
which allows students to do independent study projects on Unix workstations.
Press one of the arrows above to move around in this report.
Send comments about this publication to
pubs@cs.wisc.edu.
Mail general questions about the department
to cs@cs.wisc.edu.