Grid Computing (Sort Of) Arrives

Oracle chairman and CEO Larry Ellison kicked off the era of grid computing in feeble fashion on Tuesday, delivering a canned, droning speech that featured no demonstrations of how the technology is being used now and enough repetitive platitudes to lull at least one Oracle staffer in the crowd into a deep sleep.

When Ellison was serenaded onto the stage with a thundering display of thumping music and sweeping green lasers reminiscent of the Matrix, it seemed that a genuine event might be at hand. But what the crowd got wasn't much more compelling than a standard PowerPoint demo.

Ellison started by reminiscing about 1964, the year the first mainframe computer was invented, saying that the Beatles weren't even around yet and that the Yankees' Mickey Mantle squared off against the Dodgers' Sandy Koufax in the World Series. In fact, the Beatles landed in New York to make their legendary first appearance on the Ed Sullivan Show in February 1964, and by that April they had 14 singles in the Billboard top 100. That's a debut.

And the Yankees' opponent in the 1964 World Series was the St. Louis Cardinals.

No matter. Forty years later, Ellison says (it's 39, but OK), we have grid computing, a faster, more reliable, lower-cost alternative to mainframe-based computing models. Oracle's 10g is the flagship solution in this endeavor, layering grid-control software over a network of databases and servers to create a redundant system comprised of lower-cost devices that automatically load balance whenever traffic increases and, according to Oracle, outperforms mainframe-based systems on numerous tasks.

"Unlike mainframe systems, which have a single point of vulnerability, grid-computing systems keep running even when individual devices fail; users see no interruption at all," Ellison says. "The network redeploys resources wherever needed and automatically load balances just like a big machine does."

The most startling part of the presentation was his claim that because grid computing deploys small servers -- usually costing about $5,000 to $6,000 each and with as few as two or four lower-cost processors per box -- the solution can deliver enormous savings over mainframe-based systems, $800 per gigahertz vs. about $27,000 per gigahertz on a mainframe network. Critics like IBM are sure to decry this as voodoo economics of a mythical computer system that only exists in a vacuum; even if that's only partly true, the cost differences could still be staggering.

Ellison further promoted his vision in an invitation-only Q&A session for the press. The ability to banter one on one with reporters is a much better showcase for his flair, and the format offered a night-and-day contrast to his sluggish stage show. (It also didn't hurt that Charlie Rose and his 60 Minutes II camera crew were on hand to record the proceedings.) When challenged on whether grid computing was a lot like his network computer (NC) concept from several years back -- just another quixotic attempt to tilt at the Wintel windmill but not really useful for everyday customers and partners -- Ellison shot back. He defended the NC vision as actually having come to fruition when PC makers began realizing that scaling down their terminals into low-cost, Internet-friendly devices was what the market wanted.

"What killed the NC was the $500 PC," Ellison says.

In this way, grid computing might turn out to be quite similar. It's a broad concept that's almost purposely vague in places because it acknowledges the inefficiencies in current computing models and longs for an indefinite day when information is managed more smoothly and logically.

"The way technology works today is as absurd as if every consumer had a plane or a car built just for them," says Ellison, who almost certainly has had a plane, car and/or boat built just for him. "When I'll know if grid computing has succeeded is when I see the fundamental notion of applications changing, when we build huge information-age systems that allow us to put all our information eggs into one basket and dramatically change the quality of business."

He says a grid computing-fueled apocalypse would eliminate all but a few general purpose suppliers like Microsoft, IBM, SAP and, of course, Oracle, all supplemented by smaller specialist companies.

In the meantime, Ellison fully intends to keep selling software. Despite the lack of evidence at the unveiling, Oracle already has begun migrating customers to 10g, and the company itself is now runs on a grid model that has helped shrink its worldwide data centers from 150 to two. Ellison says VARs of all sizes will embrace the grid concept because of its economics.

"Resellers will love this thing because the initial outlay is very manageable, and it has a graceful upgrade path," he says.

We'll see. In these still-uncertain times, IT buyers and sellers can't afford to make too many purchases that seem even remotely like a gamble. But don't be surprised if five or 10 years from now Ellison is crowing about how his grid-computing vision flourished -- whether or not his company cashed in on it.