| Forward |
Metacomputing: A Reality Check
Metacomputing across the continent sounds great--on paper. But could it
really work? That's one of the questions the organizing committee of Supercomputing
'95 was asking. Supercomputing '95, the eighth annual conference for the presentation and
discussion of research in high-performance computing and communications,
took place between December 3-8, 1995 in San Diego.
At this major annual event, a Global Information and
Infrastructure (GII) testbed offered scientists a chance to transmit simulation data
remotely computed in their own numerical laboratories over the I-WAY,
a very high speed network
linking dozens of the nation's fastest computers and advanced visualization environments. This cooperative
interconnection of multiple, high performance, national ATM networks
was designed to link dozens of the country's most powerful computers and advanced
virtual environments to onsite computational resources in San Diego. These
resources included onsite virtual environments which, in some cases,
allowed viewers to interact with precomputed data, control simulations as
they ran on remote supercomputers, or even steer remote instruments.
Here's a sampling of NCSA projects that were demonstrated at Supercomputing '95:
Steps towards Realtime Astronomy
NCSA and UIUC astronomer Richard Crutcher and his colleagues attempted to show how powerful
distributed computional resources will drive the realtime astronomy in coming
decades. Their goal was to send radio signal data (gathered from a radio telescope array at Hat
Creek, California) via the I-Way across two-thirds of the country,
to NCSA for processing on the SGI Power Challenge Array.
The processed data would then be transmitted back over the network, where the scientists, using a large-screen projections
onto the "Wall," were to visualize and control the transformation of the data into images.
Computing and Visualizing Einstein's Gravitational Waves
NCSA's Relativity Group is using metacomputing tools to
solve the Einstein equations for strong gravitational fields. Solving the equations in
3-dimensions, which requires enormous amounts of computation, will help
scientists understand what happens when two black holes collide, as well as
predict the pattern of gravitational waves that would emerge. University of
Illinois physicist, Ed Seidel, and his colleagues aimed to run the computation on a group of networked
machines from a number of NSF supercomputer sites.
The goal was to enable viewers in the CAVE at San Diego
to see and interact with Einstein's gravitational waves in simulation.
Simulating and Visualizing Virtual Universes Using Distributed Multiscale Algorithms
Simulating the formation of large scale structure
in the universe is a problem requiring enormous computing power. Members
of the Grand Challenge Cosmology Consortium (GC3) used the
combined computational resources of a number of supercomputing centers,
then transmitted the data to San Diego, and displayed their "virtual" universes in the
onsite CAVE and on the Wall. The goal was to permit viewers to interactively fly
through galaxies 10,000 light years across, or even galaxy superclusters (100 million
light years in diameter).
Experimentalist's Virtual Acquisition Console
Exploring Inner Space -- at a Distance
Researchers from the UIUC and the Beckman Institute at the UIUC, and NCSA's Biological Imaging Group
used the I-WAY to control an electron microscope, an MRI Imaging Spectrometer,
and a scanning probe microscope, all located at the
Beckman Institute at the
University of Illinois. From two thousand miles away, from within the virtual environment of the CAVE at
Supercomputing'95, they managed to interact with and control their experiments.
Prototyping in Parallel
Aside from the emerging I-WAY, there are a number of other testbed projects designed to examine and further
develop the technologies of metacomputing. For example:
The National Science Foundation (NSF) and the
National Aeronautics and Space Administration (NASA)
are teaming up to evaluate the effectiveness of a variety of
scalable parallel computing systems under realistic scientific
workloads. NASA-supported R&D activities tend to be more
and mission-oriented, while NSF sponsors a broader spectrum of scientific and
computational projects. According to the rationale of the
scientists from both agencies tend to use scalable parallel computing systems
in much the same way, the case will be strengthened for the development of
truly "general purpose, " high performance computing systems, i.e.
that can be used across most applications in science or engineering.
In this testbed initiative computer scientists aim to assemble a campus-wide
"virtual computer, " a prototype of a metacomputer that will draw upon local,
regional, and national networks to connect workstations and both vector and
parallel supercomputers. The Legion project will be
Forward to the MetaCenter
Return to What Is a Metacomputer
Copyright, (c) 1995: Board of Trustees, University of
NCSA. Last modified 10/19/95.