|
August
5, 2004
By Charlotte Crystal
Greg Humphreys ponders ways to slice large problems into bite-sized
pieces.
The U.Va. assistant professor of computer
science works with
super-sized sets of data to create visual displays, a process
that requires
major computing power. But instead of looking to traditional
solutions – using
an expensive supercomputer to crunch massive amounts of data, or
simplifying the problem by reducing the amount of data involved – Humphreys
has taken a new route. He harnesses off-the-shelf PCs to work in
a “cluster” on separate pieces of a problem and later
puts the pieces back together again. It’s like solving a
giant jigsaw puzzle.
For his efforts in developing “Chromium” computer graphics
software, Humphreys and a team of collaborators recently won a
prestigious industry award, the R&D 100. The award, which recognizes
the 100 “most technologically significant new products and
processes of the year,” is bestowed annually by R&D Magazine,
a publication based in Oak Brook, Ill., that reports on applied
industrial research.
The goal of the award program, which reviews a broad array
of technological innovations each year, is to highlight
technology
that advances
and supports mankind, said Tim Studt, editor-in-chief of R&D
Magazine and program chairman of the R&D 100 Awards.
In nearly every category of analysis, Chromium received top
marks from the judges, he said.
“One judge called it ‘a killer application for cluster computing,’” Studt
said. “Another judge noted that it was ‘a major advance in the state
of the art,’ and ‘more efficient than anything else out there.’”
A
Brief History of the R&D 100 Awards
The R&D 100 Awards were established in 1963 as the
I-R 100s, in keeping with the original name of the magazine,
Industrial Research. While initially, only
U.S. companies could win, the first non-U.S. winners came along in 1965.
Over
the years, the R&D 100 Awards have recognized winning
products with such household names as Polacolor film
(1963), the flashcube (1965), the automated
teller machine (1973), the halogen lamp (1974), the fax machine (1975), the
liquid crystal display (1980), the printer (1986), the Kodak Photo CD (1991),
the Nicoderm
antismoking patch (1992), Taxol anticancer drug (1993), lab on a chip (1996),
and high-density television (1998).
SOURCE:
R&D Magazine, 2004.
|
Just a few years old, Chromium is already widely used by industry,
government agencies and researchers for a wide range of
computer graphics applications,
especially in manufacturing, defense simulations and medical science.
Humphreys, who joined U.Va.’s computer science faculty in 2002, initially
developed the “freeware,” or freely available software, while completing
his doctorate in computer graphics at Stanford University.
What Chromium does is to break down problems involving big
sets of numbers into small packets of data, like pieces
of a puzzle. The
data is sent
to a “cluster” of
computers, individual PCs that are linked together in a local-area network. Each
PC is fitted with an off-the-shelf computer graphics card that processes the
data it was sent, working only on its own little piece of the puzzle. After the
data has been converted into images, Chromium gathers the data from the individual
computers and sends it to a projector or computer screen for display. In this
way, the whole puzzle can be reassembled by using a number of projectors to display
the individual puzzle pieces side by side, creating a composite picture. The
data set is re-assembled as a visual display.
“Computer
graphics are easy to explain but hard to do, which is what attracted
me to this subfield of computer science,” Humphreys said.
One example of the software’s usefulness is seen at the Department of Energy,
which funded the Chromium research and development. Under the international nuclear
test ban treaty, the U.S. has agreed not to test its nuclear weapons. But the
government still needs to know if the stockpile works. With Chromium, the analysts
hook up a sufficient number of PCs into a cluster, process the data in pieces,
and project it to a screen as a composite image.
Another application is appreciated by video-game enthusiasts,
who like to watch video game competitions online. With Chromium
installed
on
a home computer,
a viewer can watch a game as it’s being played by professional game players,
but change the representation of art on the screen into a blueprint style so
that solid planes become open-line drawings, allowing the enthusiast to see through
the roofs and walls of buildings, observing action that otherwise would be obscured.
One of the most important applications for this technology
is scientific visualization, Humphreys said. Many scientists
today
run computer
simulations to try to understand
chemical or biological processes. But their initial results
are often in the form of huge data sets that can be difficult
to
analyze and
interpret. Using
Chromium in combination with a computer cluster, they can
create meaningful visual displays.
Chromium offers two advantages over traditional approaches
to computer graphics, Humphreys said. First, it is cost
effective. Computing
power in off-the-shelf,
personal computers has skyrocketed in recent years, making
a
lot of analytical power available for relatively little
money. Linking
them
together further
increases this capability.
Second, Chromium offers a way for researchers to create
visual displays of data in real time without having
to throw out
reams of data to
reduce the
size of
data sets so they can be manipulated quickly and easily.
Chromium is not alone in the world of cluster computing,
as custom solutions for specific applications exist,
Humphreys said. But
Chromium is one-of-a-kind
in its use of off-the-shelf hardware to run open-source
software, which is free to the public. And it’s not copyrighted — users are free to modify
it. In fact, it’s designed to be extended by its users, he said.
Along with Humphreys, the other developers named
in the R&D 100 Award for “Chromium:
Parallel, Distributed Open GL Rendering on Commodity Clusters” were Lawrence
Livermore National Laboratory, Stanford University, University of Virginia and
Tungsten Graphics Inc. The work recognized in the award was performed under contract
with the U.S. Department of Energy.
As if developing computer graphics software weren’t enough to keep him
busy, Humphreys has co-written a college textbook with Matt Pharr, a fellow graduate
student from his days at Stanford. Five years in preparation, the book, “Physically
Based Rendering: From Theory to Implementation,” is appearing this summer
as part of the Series in Interactive 3D Technology by Morgan Kaufmann, a subsidiary
of Elsevier Global Scientific and Medical Publishing.
Humphrey’s current research involves developing applications of Chromium
for use in K-12 math and science education. Through a joint Technology in Education
Program that brings together expertise from U.Va.’s School of Engineering
and Applied Science and the Curry School of Education, Humphreys is working to
enable school systems to provide educational graphics from a central location.
He’s hoping to make it easy for a school system to invest in a cluster
of computers, linked by a local-area network and served by Chromium, to transmit
visual displays for mathematics and other courses to schools throughout the system.
In the classrooms, teachers could display the
information on a variety of equipment, including
inexpensive
computer monitors,
tablet PCs,
personal digital assistants
or a single projector, depending on their
instructional needs.
Looking ahead, Humphreys believes it likely
that in 10 years or so, “most
flat surfaces,” such as walls or desktops, could be used as visual displays
in the classroom.
He sees the field of computing coming full
circle within 50 years. In the 1960s,
there was a smaller
number
of huge, central
mainframe
computers.
With the PC
revolution, there was an explosion of
small, but relatively powerful, decentralized
terminals. In the coming years, the need
for increasing power to analyze data
and create
graphics may lead
again to the
growth of
centralized
computing.
“
It doesn’t make economic sense to continue using powerful PCs; instead
we’ll probably break the coupling between display and computer,” Humphreys
said.
“What’s been holding us back has been the infrastructure of
remote delivery,” finding cost-effective ways to deliver content to computer
monitors from a distance.
And in the coming computer revolution,
Humphreys hopes his path-breaking
software will play
a leading role
in delivering
accurate, affordable
graphics to locations
near and far.
|