by Rick Tuss, Kennan Flanagan, and John Rawlins Ask most people what they know about grid computing and they'll probably mention the big scientific grids like the European DataGrid and the U.S. National Science Foundation's TeraGrid. Physicists, astronomers, and rocket scientists aren't the only ones who need lots of raw number-crunching power, though. So do chemical companies, airplane manufacturers, and banks, which is why many big companies have built their own grids.

Despite the spread of private grids, standards in this area are still in the works and there are almost no commercial grid offerings. Private grids are typically custom-built, and the high cost of keeping them running and up to date is enough to keep CFOs awake at night. At the same time, computer scientists are making rapid advances in grid computing technology. The CFO's drive to reduce costs and the scientist's drive to improve technology are combining to bring grid computing out of the lab and into the marketplace.

Some grid service providers now have solutions that can save money for companies while improving resource utilization and capabilities, without increasing their business and security risks.

Paying for the peak

The capital expenditure for a private grid can be millions, even tens of millions, of dollars a year. What's more, the technology is moving so fast that the hardware quickly becomes obsolete. Three years down the pike and it's time for new, faster servers. But the biggest headache for CFOs is that for much of the time the grids are doing nothing but sucking up money. Many companies don't need their grids 24/7, so they let them sit idle in off-peak hours.

Banks are an especially good example of this usage pattern. The peak demand for banks in New York City comes after the U.S. markets close and before the U.K. markets open. That means they have to start running their Monte Carlo simulations in the evening and finish before 4:00 Eastern Time the next morning. They can't do those calculations earlier because they're still collecting and compiling data from the day's trading.

So a bank has to pay for 100 percent of its grid's computing power even though the grid may sit idle for half of the day or more. Normally, CXOs don't want to pay 100 percent of the cost for something they use only 40 to 50 percent of the time. But to pay only for what they use, they would have to buy their computational power from someone else, which means running their data on someone else's servers.

Banks and other leading companies know that the users of an external grid are likely to include their competitors. So the idea of using outside servers sets off loud alarms about the potential for loss or disclosure of intellectual property, algorithms, and client data.

Making the cost-risk tradeoff

Security technology is way ahead of where it was just a few years ago, making even risk-averse bankers trust outside providers who have effective security measures to prevent unauthorized access. Cost is an important business driver, but so is the need to keep proprietary data secure. If security can be ensured only by paying extra, even a lot extra, those who can afford it will pay. For organizations whose budgets are tighter or whose security concerns aren't as strict, the cost-risk tradeoff will result in a decision to go outside.

CSC and other vendors have been providing fee-for-service access to high-performance computer networks for almost five years. Pay-as-you-go grid computing is similar to the mainframe timesharing services offered by CSC and other companies in the 1960s and 1970s. CSC has been providing this service to business corporations and universities since pioneering its e-HPC service (www.e-HPC.com) in 2000.

1 | 2 | 3 | next page >