Energy companies were a great source of business this quarter, DW said in the CC. This article in last month's Atlantic Monthly might hold the answer as to why: seismic imaging is very data intensive and terabytes of data must be stored to create 3D images of hundreds of miles below ground. Well written.
"The New Old Economy: Oil, Computers, and the Reinvention of the Earth" by Jonathan Rauch
theatlantic.com
snippet:
"There is nothing new about seismic imaging. The basic technology is not particularly complicated. Oil lies deep underground, trapped in the pores of rock at enormous pressures, and rock is a good conductor of sound. Knowing how fast sound travels, geologists can set off booms or pings and then analyze the returning echoes to infer the nature and location of the surface that has reflected them. Submariners have used this trick for decades; bats have used it forever. By the 1930s seismic imaging had established itself as a central technology in the oil business. Geologists would set off a boom and use mirrors to project the resulting vibrations onto photographic paper. Today the surveys use geophones wired to digital recorders, but the principle is the same.
For many years the most that computers could handle was two-dimensional imaging, which showed vertical cross sections of rock. Even 2-D created no small headache for computers. In a typical smallish 2-D survey of twenty or thirty square miles, engineers would lay out a line of hundreds or thousands of geophones and set off a bang. Then they would record the echoes, move the geophones or the sound source, and bang again. (At sea the detectors would be dragged behind boats that detonated air-gun booms every ten seconds or so.) The process would generate several million echo traces -- but such traces by themselves tell nothing about the structure of the earth. All the rest must be done by mathematical inference. The computer's job is to construct the mathematical model of the underground structures that seems most consistent with those millions of echoes; each echo must be disassembled into thousands of milliseconds, and each millisecond must be compared with many others and analyzed in several different ways before any sort of picture begins to emerge. With 2-D the picture was crude and, of course, two-dimensional.
Geologists long yearned for three-dimensional imaging, which would let them look inside the earth from any angle, but the computing requirements were staggering. With 3-D the number of possibilities that the computer needs to sort through increases by orders of magnitude. Even a small 3-D survey might easily generate 200 gigabytes of data, and all the data points would then need to be analyzed from every direction simultaneously. In 1975, 3-D seismic imaging came into its first commercial use, but it was too slow and expensive to do anyone much good.
That is what the processing revolution changed. From 1985 to 1995 the computing time needed to process a square kilometer's worth of data fell from 800 minutes to ten minutes. From 1980 to 1990 the cost of analyzing a fifty-square-mile survey fell from $8 million to $1 million. Now it is more like $90,000.
In 1989, according to the U.S. Department of Energy, only five percent of the wells drilled in the Gulf of Mexico made use of 3-D seismic data; by 1996 the figure was 80 percent, and the surveys themselves were more sophisticated. Today that smallish twenty- or thirty-square-mile survey might easily involve 25,000 geophones, 2,000 echo traces per shot, and 30,000 shots -- a total of more than half a million floppy disks' worth of data, or more than 800 gigabytes, of which each byte must be compared and reconciled with thousands or millions of others. Over the past decade seismic analysis has increased its data-gathering and data-processing requirements by an order of magnitude every five years, and no end is in sight. Schlumberger, a big oil-technology company, recently announced a higher-resolution technology, called Q, that gathers several times as much data as anything before and pushes computers proportionately harder.
The easiest way to appreciate the results is to go look at them. A good place to do that is at Magic Earth, in Houston. Magic Earth is what is known as a 3-D seismic visualization facility, one of many that have appeared in the past three years. "Engineers sit in front of this," Michael Zeitlin, Magic Earth's president and CEO, told me one day last spring when I stopped by for a demonstration, "and have an emotional response."
Zeitlin has been in the oil business for twenty of his forty-two years. He is originally from New York City, but his manner is all West Coast. We were sitting in a glass cube of a building, in whose center stood a twenty-five-foot-high metal pod, like a shiny mushroom cap. Zeitlin talked with the rat-a-tat energy of an entertainment-industry executive, which he could almost be. Hollywood uses digital imaging and supercomputers to bring dinosaurs and typhoons and starships to life; Zeitlin does the same thing for structures deep in the planet, structures that will never be seen by any human eye.
Vision, Zeitlin explained, requires a great deal of bandwidth. To generate the illusion of motion requires fifteen to twenty frames a second. In the 1980s geologists' computers couldn't calculate that fast; in the 1990s they closed the gap, and then some. High-end computers can now take terabytes of seismic data and create vividly colored three-dimensional images of them; and they can generate those images at a rate of up to sixty a second -- fast enough to fool the brain into thinking that it is panning through the interior of the planet, seeing formations that are miles underground in much the same way one might see the Grand Canyon from a moving helicopter." ...more |
