Although the installation of a new supercomputer at the National Oceanic and Atmospheric Administration (NOAA) earlier this year may seem worlds away from a System i quietly running a customer database, the two are related. They share hardware, after all, because the supercomputer is a gang of System p servers. They also have the same manufacturer, IBM.

And that, says IBM deep computing marketing manager Herb Schultz, makes the latest NOAA supercomputer of more than academic interest for System i users. "These large systems tend to go into the bigger, more technologically challenging customers running really hard applications, and the things we've learned from that go back into refinement and the development of the product, which then benefit a broader audience," Schultz says.

In the case of the NOAA supercomputer, the system is indeed large. It has 160 p575 servers, each with 16 1.9 gigahertz POWER5+ processors. It can do 14 trillion calculations per second, or 14 teraflops, and is ranked 36th among the top 500 fastest computers worldwide.

According to Schultz, IBM's Blue Gene is currently number one on that list, and it, too, shares something with the NOAA supercomputer and System i servers. "It's also built in Rochester," he says.

The new NOAA system is more than a threefold increase in terms of computing power over its predecessor. In addition to its processing muscle, the supercomputer has 160 terabytes of storage.

Activated in January, the supercomputer is going to make the weatherman's job easier. To paraphrase Bob Dylan, you might not need a supercomputer to know which way the wind blows, but you need one when you're trying to figure out which way the wind will be blowing a week from now.

Working with satellite data, the supercomputer will help with both weather and climate modeling. In particular, it will crunch data from a series of six observation satellites launched in 2006 and thereby provide a better understanding of the jet stream and related storm systems.

There's even hope that the more highly advanced models produced by the new system will help with forecasting hurricane intensity, a problem that meteorologists haven't yet cracked.

When doing weather forecasting and climate modeling, a large amount of data has to be handled, with many different variables interacting. Those requirements drive both the processing power and the storage needed.

In addition to being technological proving beds, supercomputers also impact System i servers in a more mundane way. Big systems like the NOAA supercomputer are built out of a lot of components, ranging from thousands of processors to a similar number of fans and on to a much larger number of connections. Since expensive supercomputers are expected to run reliably, all of those parts have to be assembled correctly. Just one failure during manufacturing or later can be catastrophic for the finished product. "Sometimes if one thing breaks, the whole system can be rendered unusable," Schultz says.

For that reason, he adds, IBM has developed a manufacturing capability to hopefully avoid, or at the very least minimize, such problems. That expertise clearly benefits big systems. However, even small systems, such as a humble System i, that go through the same manufacturing line will be more reliable as a result.

This isn't the first supercomputer IBM has delivered to NOAA. The company is halfway through a nine-year contract. So far, an upgraded supercomputer has been delivered every two to three years, an interval driven by technology advances and the time required to bring a new machine online. As for why IBM originally got the contract, NOAA spokesperson Carmeyia Gillis says it's very straightforward. "The p system outperformed on our benchmarks. Thus, this is the reason why it was selected."