In most applications, the power of modern processors has almost eliminated concerns about computational performance. But according to Srinivas Koushik, enterprise CTO of Nationwide, financial companies don’t have it quite so easy. “Insurance and financial services companies — one thing they have a lot of is data,” Koushik says. “We’re constantly mining for information.”
One of the big computing challenges Nationwide faces is the development of pricing strategies for its new products. To accurately predict the success of offerings in today’s markets, Nationwide relies on stochastic analysis, a process that applies random variables to a statistical model to simulate thousands of hypothetical scenarios and their potential outcomes.
“Those are pretty computing-intensive calculations,” Koushik explains. “Many of these calculations on a dedicated server tend to run for days. Not hours and not minutes. They run for days.”
Rather than throw money at the problem, Ron Santacroce, senior director of life actuary at Nationwide Financial, suggested a whole new paradigm. In 2004, Nationwide partnered with grid computing vendor DataSynapse to develop a way to distribute Nationwide’s computational workloads across its existing infrastructure — including not just servers, but desktop workstations as well.
“Grid computing was an ideal fit, because we ended up putting in what some people call a ‘desktop scavenging grid,’ ” Koushik explains. “I’ve got 50 or 100 desktops connected into the network that’s always plugged in. Why not scavenge the excess capacity and excess computing power that’s sitting on that to be able to run these calculations?”
The true ROI for Nationwide’s grid efforts is difficult to estimate. But if time is money, Koushik and Santacroce already have hard numbers in hand. With just 200 workstations added to the grid, the time to calculate one analysis has been slashed from 36 days to 26 hours. Another report that might have taken 27 days now finishes in just eight hours. And Nationwide is adding more workload to the grid all the time. “Demand has quickly exceeded capacity, even at 200 engines. Next year we’re thinking about running 500 engines,” Santacroce says.
According to Tony Bishop, chief business architect at DataSynapse, Nationwide’s efforts are blazing a trail for the growing number of businesses that will adopt grid computing in the coming months. “Nationwide is what I would define as a visionary customer,” he says. “They’re 12 to 24 months ahead of the cycle, where you’re going to see the mass explosion of this across a predominant amount of business.”
If that means grid computing has given Nationwide a one- or two-year lead on competitors, then Koushik and Santacroce should sleep a lot easier.