Geophysics researchers increase computing resources

A new computing cluster will allow Virginia Tech's geoscientists to see the earth’s mantle in greater detail than ever before and answer such questions as, "Just how big is the plume from the earth's core that fuels Hawaiian volcanoes?"

Scott King, professor of geophysics, and Ying Zhou, assistant professor of geophysics, have built a new computational cluster that is 1,000 times more powerful than a typical home PC, said Kevin Shinpaugh, director of research and cluster computing at Virginia Tech.

The system consists of 96 Dell 1950 servers; each node has two Intel Clovertown (quad-core) processors and includes 12 GB of ram. The total system has 768 cores, Shinpaugh said. The system interconnect is double-data-rate (DDR) InfiniBand with a 144 port DDR QLogic switch.

The High-Performance Earth Simulation System (HESS) has eight times the capacity of the 96-processor cluster the researchers are presently working with, King said.

“The interior of the earth is an inaccessible place and we have to study it using indirect observations and numerical models,” said King. “With the new cluster, we will be able to do jobs eight-times faster. Things that took two weeks will take two days. More significantly, we will be able to do bigger jobs – with more information. We can put detail and physics into a model that we couldn’t before,” he said.

More information on King’s research may be found online.

For example, Zhou’s research depends upon information from seismic tomography, which uses the energy of earthquake waves to create three-dimensional images of the Earth’s interior from 2,800 kilometers (1,740 miles) down to the surface. “We have not been able to make use of all the information that was collected because of limited computing capacity,” she said.

The speed of the waves is altered by the temperature structure in the mantle, but models have had to discard details of velocity (temperature) variations. “Previous models are not able to distinguish between small scale hot materials and large scale anomalies accurately,” Zhou said. Since waves travel more slowly through a heated rock, “If in the previous mantle models, you see a 2000 kilometer diameter rock hotter than surrounding rocks, the real size of that hot stuff is probably only 200 km in diameter. The previous models can only ‘see’ large scale anomalies,” she said. “Also, the mantel is moving because some parts are hotter than other parts; so seismic velocities are different.

Zhou is trying to understand seismic velocity structure, temperature structure, and how material moves in the mantel. “With the new cluster, I hope to be able to translate seismic data into velocity and temperature structures in the crust and mantle, and probably the inner core. We will be able to image the earth's interior at high resolution – and learn the dynamic processes in the Earth,” she said.

More information on Zhou’s research may be found online.

King’s research focuses on the dynamics and evolution of the interior of the terrestrial planets. “The process by which the Earth cools is the driving force behind most tectonic and volcanic activity. Mantle convection and plate motions are part of a linked system. For many of the problems that I am interested in, it is important that we model the full spherical shell of the Earth's mantle,” he said. “A lot of progress has been made with two-dimension or three-dimension models of regions of the Earth, but to get a better understanding of how the Earth works, we need to understand how different features interact with each other. To do this we need lots of computer power,” King said.

“I use the computer as my laboratory and computer models as tools for hypothesis testing,” he said. “Now we will be able to ask more ‘what if’ questions with our models. For example, rocks deform in a certain way under pressure. If I change the rule for how that happens, I may end up with something frozen like on Mars, which provides a chance to understand how Earth and Mars are different.”

The HESS cluster -- named after geophysicist Harry Hess, whose work was instrumental in the development of plate tectonics, gives Virginia Tech’s geophysics research lab resources beyond most such labs. “What is unique about Virginia Tech is the computing capacity per researcher,” said King. “This Dell cluster puts us in a position to consider problems others cannot, such as resolving the plume structure of Hawaii volcanoes.”

Volcanic activity usually happens at the edge of the plates that move atop the less rigid mantle, but Hawaii is in the middle of the Pacific plate. “The idea is that a plume from the earth’s core pierce’s the plate,” said King. “Seismologists tell us the opening is no wider than 100 kilometers (62 miles). But we could not resolve that with previous generations of models. Now we will be able to do the necessary fluid dynamics.”

In the process of defining their needs, the Virginia Tech researchers consulted Jeroen Tromp, professor of geophysics and director of the seismological laboratory at California Institute of Technology. “He recommended Dell as they were working with Dell in upgrading their cluster to a quad-core system, the type of system Virginia Tech is interested in,” said Zhou, who worked at Caltech and is familiar with the performance of the Dell system for seismological applications.

“We collaborate with Caltech, so our codes will work back and forth,” said King. “And Dell worked with us on the changes we needed.”

HESS will be located at the Advanced Research Computing facility in the Andrews Building. “It doesn’t matter where it is. We can access it from our desktop computers,” King said.