| ACES
and Clusters Revving Up Environmental Research
New
computing initiatives to accelerate DRI and UCCSN science programs
| 
|
| Modeling
reveals reality. Dr. Vanda Grubisic´
has been working with computer models to depict the flow of
air over the European Alps, as indicated in the graphic behind
her. The future ACES capabilities and new high performance computing
assets at DRI will allow her to extend this field of research
and conduct modeling and simulation operations in a matter of
hours where they previously required a day or more. (Photo
by John Doherty; graphic by Vanda Grubisic´) |
With major new funding
from the National Science Foundation and the state of Nevada, and
several important computer-based research grants from federal agencies,
scientists from the Desert Research Institute and the University
of Nevada campuses in Reno and Las Vegas are about to leap to an
entirely new level of research and discovery.
The springboard for these
advances is a $3.4 million Advanced Computing in Environmental Sciences
(ACES) Program, which will involve the development of a sophisticated
research grid among the three University and Community College System
of Nevada (UCCSN) research campuses—the University of Nevada,
Reno (UNR), the University of Nevada, Las Vegas (UNLV), and DRI—during
the next three years, says DRI Vice President for Research Dr. James
Coleman. Coleman says ACES, funded under the National Science Foundation
(NSF) Experimental Program to Stimulate Competitive Research (EPSCoR),
will provide a high-powered research computing grid in Nevada with
a link to the new “Internet 2” national research network.
 “The
main mission of ACES,” says Dr. Vanda Grubisic´, the
lead UCCSN scientist for ACES and an assistant research professor
in DRI’s Division of Atmospheric Sciences, “is to elevate
environmental science research and education in Nevada to a new
level of excellence. We will do that by aligning it with NSF’s
strategic direction of increasing the application of information
technology (IT) advances across science and engineering.”
Grubisic´ says the
ACES program is employing an ambitious two-part strategy. “First,
we will create the Nevada Environmental Computing Grid (NECG), a
statewide, distributed computational and collaborative infrastructure,
by acquiring new equipment, making several strategic hires, and
integrating existing resources. It should be emphasized that NECG
will be built upon the existing high-bandwidth communication infrastructure
of the Nevada Research Network. An important part of the NECG building
effort will be to establish a unifying management approach and transparent
access and use policies, with dedicated application level support
for researchers. Second, our intent is to foster new interdisciplinary
IT research at UCCSN campuses, and for that it is important to stimulate
interaction of computer scientists and environmental scientists
through seminars, graduate level courses, tutorials, and workshops,
all accompanied by online reports for the widest possible dissemination.”
ACES will expand capabilities
for both Nevada environmental and computer science researchers in
an impressive range of topics. While the former will be able to
apply advanced numerical modeling and other data-intensive research
techniques for studies on groundwater flow, hydrochemistry, floods,
atmospheric and air quality modeling, climate data analysis and
modeling, and geochemistry modeling, along with applications of
geographic information systems (GIS) and remote sensing, the latter
will be able to advance their research and development of algorithms
for high-performance computing and scientific visualization.
|
Difficult to perceive.
Three-dimensional models of air flow over mountains and rough
terrain can improve climate analysis, weather prediction,
and land use decisions. (Graphic by Vanda Grubisic´.) |
Grubisic´ said
part of NECG will be AccessGrid Nodes—sophisticated multimedia
environments—at UNR, DRI, and UNLV that will accommodate high-end
group-to-group electronic collaboration, scientific visualization,
and other interactive online data applications. The ACES funding
will also provide for one high-end computational server. Located
at DRI’s Northern Nevada Science Center in Reno, the platform
will be accessible by scientists throughout the ACES network. This
is also true for another ACES expansion, a computer cluster—a
group of personal computers (PCs) interconnected by a special networking
system—to be located at UNLV’s National Supercomputing
Center for Energy and the Environment in Las Vegas. In addition
to this, ACES plans call for integration of the existing computational
resources at all three campuses.
DRI will be recruiting
a dedicated system administrator and a scientific applications engineer
to staff the ACES system, Grubisic´ noted. “The ACES
staff will be very busy with responsibilities for implementing and
maintaining the system, implementing software for parallel computation
and scientific visualization, and providing programming support.”
Campuses, on the other hand, are looking into recruiting at least
three new faculty specializing in computer science and applied mathematics.
The ACES staff will also conduct tutorials for users throughout
the UCCSN and facilitate the integration of a variety of types of
computers into the NECG.
While the ACES network
will greatly facilitate scientific interaction within the UCCSN,
for straight scientific “heavy lifting,” DRI is turning
to new in-house computer clusters for computational firepower.
| 
Growth in clusters.
Dr. Boro Grubisic´ is building several computer clusters
to aid DRI research in atmospheric processes and range fire
prevention and management. This 64-processor cluster will
be followed by another 128-processor system. (Photo by
John Doherty) |
Grubisic´ says
clusters are a relatively recent development in parallel computing
that use multiple off-the-shelf computer processors configured together
by new networking equipment and software. Clusters have greatly
reduced the cost of running extremely large and complex simulations
of natural phenomena. Across the scientific community, researchers
are turning to clusters as the cost-effective solution to the massive
number crunching requirements of mathematical models that describe
and simulate physical reality and environmental processes. DRI’s
scientists have obtained major new research funding to develop and
operate clusters that will improve the Institute’s environmental
research capabilities. “We
can now accomplish, for several hundred thousand dollars of computing
investment, the level of ‘number crunching’ that would
have required several million dollars a few years ago,” says
Grubisic´, “and with this approach we even have the
option of expanding the system in the future, rather than having
to buy an entire new computer system.”
With a $280,000 major
instrumentation grant from the National Science Foundation, Grubisic´
and her physicist husband, Dr. Boro Grubisic´, are assembling
a 64-processor cluster at DRI, called the “Sierra Cluster.”
The new cluster supports Grubisic´’s continuing research
involving data-gathering flights in the vicinity of the European
Alps to uncover how the mountain-generated airflow patterns influence
local and regional weather (See “Flying in the Face of Stormy
Weather,” DRI News, Fall 2000; http://newsletter.dri.edu/2000/fall/vanda.html).
Grubisic´ has also
recently been awarded more than $1 million in new research funding
for several other large projects specifically targeted for the capabilities
of her new computer cluster. Besides studies of airflow phenomena,
the work includes basic research on orographic clouds and precipitation.
As soon as Boro Grubisic´
finishes his wife’s 64-processor cluster, he will begin work
on another, even bigger, cluster designed to aid in wildfire management
and fire-smoke air quality prediction. The larger cluster, totaling
64 dual processors, or 128 powerful linked computer chips, is expected
to become operational early in 2003. It is the heart of a new project
under the direction of Dr. Tim Brown.
| 
|
| Proactive
fire management and support.
Dr. Tim Brown, director of DRI’s CEFA program, with a
monitor showing a CEFA map graphic generated from about 500,000
federal wildland fire occurrence reports that shows the percentage
of fire starts in the West attributed to lightning. The lightning
starts increase in ten percent increments from dark blue up
through green, yellow, orange, and red. Orange and red areas
imply fires mostly caused by lightning, while blue and green
areas imply mostly human-caused fires. (Photo by John Doherty) |
A DRI climatologist,
Brown is director of DRI’s Climate, Ecosystem, and Fire Applications
(CEFA) Program which performs applied research and develops informational
products related to climate, weather, fire, and natural resources
(Website: http://www.cefa.dri.edu/).
His cluster will run predictive models for CEFA’s Operational
Forecast Facility (COFF) and serve a broad customer base of 13 local,
state, and federal agencies in Nevada and California that have signed
on for an initial five-year program totaling nearly $2 million.
“These agencies
are constantly making decisions and policies related to wildfire
suppression, prescribed fire and fire use, and smoke related to
fire and air quality,” Brown says. “This process requires
accurate predictive meteorological information, which can only be
acquired from computer intensive physics based numerical models.”
Brown’s computer
cluster will operate a model incorporating the National Weather
Service’s massive daily generation of weather data with the
objective of producing predictions of fire-relevant meteorology
on a daily basis.
Brown said the system
will operate using meteorological data taken from points as close
together as 4 kilometer intervals, which may sound like a long distance,
but is currently considered high resolution for this type of analysis.
In some cases, such as fire behavior analyses, those data points
will be as close together as 500 meters. “We anticipate that,
because of the high resolution of these predictive models, the forecast
products will also prove valuable to wildfire managers on major
firefighting efforts.”
“Cluster”
may seem an inelegant term to describe the sleek arrays of softly
humming computer processors coming online around DRI, but, for researchers
challenged to describe the real world in real time, it’s the
easiest way to “bring it all together.”
–John Doherty
|
