As Published In November 2005

Grid Gains Traction
By Jeff Erickson

Grid computing is driving revenue and savings in the enterprise.

Grid computing—the linking of many servers to create one big computing resource that provides processing power where it's needed—has emerged from the research centers of its inception and is gaining momentum in the enterprise. Industry groups such as the Enterprise Grid Alliance and the Global Grid Forum continue to pick up big-name members, from NASA to Intel, and the press continues the buzz about a "paradigm change." But what's it like when a CIO and his team take the plunge and put core revenue-generating systems onto the grid?

Profit talked to Kevin Vasconi, CIO of R.L. Polk, a worldwide provider of automotive industry intelligence and marketing solutions, about the business and technical realities of his company's recent enterprise grid implementation.

PROFIT: What made you give grid the green light?

VASCONI: A lot has been written about grid in the media. All those claims are what originally got us thinking about it. Then we started to see established companies such as Oracle embrace the philosophy and start to turn the hype into real products. Finally, I was at a CIO summit and saw my peers starting to move grid into action. The CIO of Electronic Arts talked about launching The Sims Online on a massive grid using commodity servers. That's when we decided to look at it for ourselves.

PROFIT: What did you discover?

VASCONI: It's a pretty cut-and-dried value proposition. It's reducing the total cost of ownership. We're conservatively estimating 30 percent savings from moving a clustered UNIX platform onto an Intel- and Linux-based grid using Oracle Database 10g. When we migrate other parts of the business off mainframes, our savings will be even higher. We predict at least a 40 percent return on investment.

PROFIT: So you did it for the infrastructure savings?

Grid 101 Download pdf

VASCONI: We would have done it even if that had been the only benefit. But we're also picking up improvements in our service-level agreements. Grid computing is more reliable, because there aren't as many single points of failure.

Another benefit is in capacity planning. With grid computing, we can approach it with less risk to the business, because when we need more capacity we aren't buying another US$300,000 mainframe processor or throwing another Sun StorEdge 8600 at it. Even slapping a couple of processors into a Sun or HP UNIX box is still a pretty expensive proposition. Contrast this with an Intel-based grid, where we're talking literally about throwing a couple of PCs into the mix.

PROFIT: How should a CIO sell the management team on the idea of grid computing?

VASCONI: Stress the business-ROI side. Right now, people like to talk about grid computing because it's technology fashion, on the forefront. Yes, it's really cool and really neat, and we technologists love to work on stuff like that, but I would tell the IT audience to resist talking up that part of it. Talk about the business impact of grid technology and what it can mean to the total cost of ownership—to your company's bottom line.

My other recommendation would be to cut your teeth on a pilot project that is doable and has impact. There are plenty of high-impact business problems that are tangible to both the IT and business communities.

PROFIT: What about the naysayers?

VASCONI: The biggest issue we struggled with was getting the business to understand why this was an acceptable level of risk and how the new computing paradigm has business-related benefits. Anytime there's a paradigm shift, you're going to get resistance, and a true change is what we're talking about here: This is more than just technology change; this is changing the way people think about technology, the way they implement it, and the way they budget for it. By the time you've overcome the resisters and the technical naysayers, the difficulties in migration are probably two or three times as difficult as a migration within the same technology.

PROFIT: How do you prepare for technical pitfalls?

VASCONI: You're going to uncover problems in a migration such as this, because you're not affecting just a technology; you're also affecting a way of doing business. My advice is to expect the unexpected. Do as much contingency planning up front as you can, and try to build a team nimble enough to deal with unpredictable contingencies.

PROFIT: Can you give an example of a technical gotcha?

VASCONI: The hype is that when you use blades [blade server technology that allows you to reduce the cost of hardware and increase density], your rack space goes down to nothing. We started implementing blades in an effort to lower our data center costs, given that we pay by the square foot and for all the accompanying services. But once we started designing, we discovered that the huge amount of heat generated by such a compact platform meant that we couldn't put anything near it. We had to rethink the blade density and didn't end up saving the space we thought we would.

For More Information WATCH demos, read white papers, and learn about the steps for getting on the grid LEARN more about the Enterprise Grid Alliance

PROFIT: Any unforeseen benefits?

VASCONI: The technology lift from grid computing is a wonderful thing and worth the trip, but if that's all you consider, you're leaving a lot of money on the table. You need to look at how your whole organization is going to take advantage of the grid. You just can't keep all your old business processes if you're going to get the full benefit from it. During our implementation, we would uncover a potential business benefit and be forced to decide whether or not we were willing to go get it. That's probably where we've had the most pleasant surprises.

PROFIT: Would you do it again?

VASCONI: Based on what we saw in the initial implementation, we're now on a mission to move the majority of our applications and infrastructure onto a grid platform, because we've seen the results. Whatever the issues and pain that need to be overcome to get on the grid, the offsetting business value more than compensates the organization for going through the trip.

Grid and Big ROI? As one of the world's largest grid implementations, the Oracle Data Center in Austin, Texas, is a showcase for the ways grid computing saves money. "Sure, we save money in the ways you would immediately expect," says Bob Thome, Oracle's product manager for distributed database product management, "but we save money in subtler ways as well." Immediate Savings In its grid configuration, Oracle saves money on equipment, by standardizing on low-cost servers with Intel processors running Linux. Oracle also achieves savings by centralizing its computing resources. "Our Austin data center is used by Oracle production and development applications all over the globe," Thome says. "If we maintained independent local data centers, we would need to pay teams of administrators in all those places." Oracle also saves money by pooling computing resources and sharing them across applications. Long-Term ROI "You can get an often overlooked savings from grid computing, in that you can get better-quality work done in less time, because you've got more CPUs available," says Thome. "Many of our enterprise customers will find that on-demand access to vast computing resources provides their greatest opportunity for long-term returns. "For example, Oracle programmers have thousands of tests they can run against a new line of code," Thome says. "But a single test against Oracle's code base can eat up an hour on a single workstation, so programmers might choose only the three tests most likely to find any problems. Only later do they find that the new code has broken something they hadn't tested," says Thome, "which can cripple an application and cost the company big." Tapping the Grid With the ability to dynamically draw immense computing power from Oracle's Austin Data Center grid, a programmer can now run all the tests—finding problems before they create trouble. "More information, better decisions, and more uptime," Thome says, "are big benefits of the grid." A similar example can be found in large financial institutions that perform complex analysis on their data to mitigate risk. "There are instances in which financial institutions have networked their CPUs into a grid to turn several days' worth of algorithm crunching into an overnight operation," Thome says. "They have leveraged grid computing to get faster results and make better decisions." The same dynamic works for order-processing systems during holiday spikes or sudden peaks in usage of a corporate Web site. "The benefits of power-on-demand that we've seen working for Oracle in the Austin Data Center are starting to appear for companies in all sorts of industries," says Thome, "and the trend has only just begun."