Author: Jeff Bronfeld

What exactly makes a computer super? In our industry we are used to measuring things in Kva and Mva, horse power, voltage, PSI, degrees Celsius. But our highly integrated power systems are depending more and more on computational "power", which is measured in flops. (Hey, don't laugh, we measure things in darafs, a unit of electrical elastance, the reciprocal of farads.) Flops are floating point operations per second and are somewhat similar to instructions per second, especially for computer tasks requiring lots of calculations. Flops are a measure of a computer's ability to compute, which is, after all, what you want it to do.

As you would expect, there are

Megaflops (Mflops = 1,000,000 flops),
Gigaflops (Gflops = 1,000 Mflops),
Teraflops (Tflops = 1,000 Gflops)

So let's get some perspective. The CPU of a typical desk top personal computer today, say a Pentium 4, would have a rating of around 10 Gflops, somewhat less when the entire computer is taken into account.

So do you have a need for FLOPS? Do you want to calculate state estimation with real time measured data, and have your power system assess its own threats and take corrective action? Do you want your protective relays to become the eyes and ears of an intelligent power grid that predicts its own performance and recommends its own enhancements, or reconfigures itself based on expected load patterns, or weather conditions?

You need flops? You got flops!

The web site www.top500.org lists the 500 super-est supercomputer sites. From that site we see:
On the new list, the IBM BlueGene/L system, installed at Department of Energy's Lawrence Livermore National Laboratory (LLNL), retains the No. 1 spot with a Linpack performance of 280.6 teraflops (trillions of calculations per second, or Tflops).

The new No. 2 systems is Sandia National Laboratories' Cray Red Storm supercomputer, only the second system ever to be recorded to exceed the 100 Tflops mark with 101.4 Tflops. The initial Red Storm system was ranked No. 9 in the last listing.

Slipping to No. 3 from No. 2 last June is the IBM eServer Blue Gene Solution system, installed at IBM's Thomas Watson Research Center with 91.20 Tflops Linpack performance.

The new No. 5 is the largest system in Europe, an IBM JS21 cluster installed at the Barcelona Supercomputing Center. The system reached 62.63 Tflops.

The ranking is constantly changing, as technology moves on. DOE, for example, is looking at the next generation of supercomputers, capable of Petaflops (1000 Teraflops) operation.

What are all these super computers doing? Lawrence Livermore National Laboratory has a computer named "Purple", which calculates at about 100 Tflops. The computer is calculating the explosion of a nuclear weapon, from "button to bang"; a task that in 1994 the world's fastest computer would take 6000 years to calculate. NEC's earth simulator is attempting to predict the weather 100 years into the future, based on satellite measurements today.

Looking at these supercomputers I am reminded of a phrase I learned in engineering school; "Don't use a pile driver when a tack hammer will do". Luckily advances in computing technology trickle down to all levels. There are commercially available "personal supercomputers", or so called PSCs. Tyan offers one such PSC, capable of 256 Gflops, for about $20,000, and it draws about 15 amps current. What is that equivalent to - say a clothes dryer or a very large refrigerator?

We used to let the tools available limit our imagination. In the world of computers at least, things appear reversed - imagine what you want done, the computers will be there to do it.

Photo caption: BlueGene/L, Terascale Simulation Facility (TSF), Photo courtesy of LLNL