Visualization Tools Illuminate Complex Phenomena
Visually studying the effects of pressure and temperature changes on fluid flow in petroleum-rich stratigraphic layers in the Louisiana Basin is an example of a leading-edge research project that, in many ways, is no different from the hundreds of others underway at the Center for Theory and Simulation in Science and Engineering at Cornell University -- popularly known as the Cornell Theory Center. But this and other chosen activities employ state-of-the-art supercomputing visualization tools to help the investigator comprehend complex physical phenomena and their underlying principles. POWER System Visualization methodology is "as applicable to industrial design and engineering as it is to scientific research," points out Mal Kalos, director of the Cornell Theory Center. At the Center, this capability is derived from a potent hardware/software visualization environment comprised of the 32-processor POWER Visualization System (PVS), built by IBM to address high-end visualization requirements, and Visualization Data Explorer (DX) software, also developed by IBM, which boasts an approach to applications programming that encourages even computer-shy scientists to use the system. The PVS/DX system is now part of the formidable lineup in the Center's Cornell National Supercomputer Facility, a world-class supercomputing environment. Beside the PVS, the facility houses an IBM E5/9000 Model 900 vector mainframe and a scalable cluster of 32 IBM RISC System/6000 workstations, among other high-powered machines. Geological Research Take, for example, the Global Basin Research Network (GBRN), a consortium of major oil companies and universities led by Cornell and Columbia, that is the leader in geological research. Research concerns a 30 x 6 km section of the Louisiana Basin, an oil-rich area in the Gulf of Mexico, and hopes to understand further the dynamic conditions necessary for the production of oil. The ultimate goal, of course, is to enable the companies to extrapolate their findings to predict where oil most likely can be found elsewhere around the globe. Over the course of many years, individual members of the consortium have collected seismic data on this portion of the sea floor, but no one had ever seen all the data put together in a meaningful way. Enter the PVS/DX System at the Cornell Theory Center. Representing the GBRN consortium, scientists Roger Anderson of Columbia University and Larry Cathles of Cornell put the supercomputing visualization system to work, first to obtain static views of the data, then to create simulations. The motion of fluid (oil) through the irregular fault structure, derived from computational fluid dynamics, was superimposed on the dataset where an observer could simulate "walking around" or "diving into" the data structure to observe the flow in process. In the investigation, the PVS provided the lightning-fast calculation speed needed to create the images, and the visualization software supplied the brain power. DX orchestrated the rendering of images showing the strata and the fluid flow. It managed the large datasets, actually rendered the images and generated the animation frames. Furthermore, all of this could be visualized for a time span of many years because of the historical data available. DX enables the scientist to view a PVS-generated animation, then change the point of view or lighting characteristics of the data, and to repeat the scenario again for further refinement. When the executing platform is fast enough, this interactive orchestration happens in fractions of a second. "Such rapid feedback makes the machine an extension of the scientist's mind, "comments Siegel. "It affords insights not available otherwise." Long-Standing Relationship The primary reason that the Center acquired PVS was the potential it offered for dramatic advances in visualizing sophisticated science. Simply stated, "PVS is the fastest supercomputer for visualization available -- the fastest image rendering engine focused on scientific problems," Siegel asserts. Although there are no formal plans to expand the PVS capability, additional computer suites are likely to show up as the escalating demand causes the existing PVS to become saturated. In addition, the communication bandwidth for remote users will eventually be upgraded to data transmission rates close to the 100Mps that Center-based local users currently enjoy. As for the Data Explorer, Cornell researchers are helping IBM develop additional program modules that add functionality to the software. They have just launched a national DX repository, allowing a large audience to contribute and use a variety of Data Explorer modules.