Can Ellison Win With Grid?

In fact, the way Oracle is hyping grid is remarkably similar to how it proselytized the network computer (NC) several years ago. Back then, Oracle chairman Larry Ellison traveled far and wide to promote the NC as the logical evolution in desktop computing and,in no small coincidence,the perfect antidote to the ruthless domination of Windows PCs.

We all know that the NC didn't succeed exactly the way Ellison envisioned, but it did trigger a movement among PC makers to streamline desktops into Internet-optimized terminals that cost next to nothing. Now, Oracle hopes grid computing will similarly revamp the industry, resulting in systems that automatically share and reallocate resources across a decentralized infrastructure, cost less to implement and upgrade, and make it easy to plug new devices into the network.

While there are few doubts about those benefits, the trick for solution providers will be to judge how quickly to migrate themselves and their customers to the appropriate setup.

"Two, three, five years from now, we'll see applications like CRM, ERP and supply-chain management all running on a grid infrastructure, but to roll out these applications you have to think big yet start small," says Ian Baird, chief business architect at Toronto-based VAR Platform Computing.

id
unit-1659132512259
type
Sponsored post

Grid In the Making

A cousin,and competitor,to the ideas of on-demand and utility-based computing, grid is the partial creation of the Globus Alliance, a consortium of international academic computer-science laboratories that has created test beds, applications and toolkits that help promote the framework. Ellison also has cited Digital Equipment's old database-clustering technology as a contributor to Oracle's version of grid.

"We started looking into the clustering question 14 years ago; we never got it fully working, but the way everything else worked got better," he says.

Today, Oracle has fully adopted the idea for its 10g database, putting its money where its mouth is internally by adapting its own networks to the grid concept. In the process, it reduced its 150 worldwide data centers to two.

But the curious thing about the OracleWorld exhibition,which Oracle promoted as the grand unveiling of 10g,was its lack of many demonstrable, real-world examples of Oracle-sponsored grid technology at work today. However, the company did parade out some customers for breakout press sessions to talk about grid's low cost and ease of use.

"We hope to use 10g to ease deployment, management and clustering of our data for our worldwide computing grid project," says Jamie Shiers, database group leader for the IT division of the European Organization for Nuclear Research (CERN), which uses grid for the data-storage requirements of a massive nuclear particle-exchange project.

Similarly, medical research publisher BioMed Central is using 10g to reduce production and data-storage costs and to make data available on demand. And Ordinance Survey, a British mapping organization, uses 10g to store more than 430 million topographic features in one database.

"We've found no other database that can meet our requirements for easy and open access," says Ed Parsons, Ordinance Survey's CTO. "It is dramatically easier for us to maintain our data using 10g."

What these customers have in common is obvious: They all tend to be concentrated abroad and in academia or related sectors,not the sweetest market spot for most North American VARs. Moreover, in late September, Oracle announced it will give away 10g to its existing customers, thus undercutting a potential revenue stream for certain VARs, though they still will be able to peddle integration services to 10g customers. The rationale is by forgoing some revenue now and saving its existing customers money, the company will foster loyalty that will pay off later.

Those VARs that are deploying grid solutions now are primarily ones who have spent years working on the concept. Platform Computing, for example, has been at it so long,about 12 years,that last year Oracle approached it to help grid-enable Oracle's technology. Platform now resells Oracle's grid offerings to customers and helps other VARs adopt grid solutions.

"Oracle has shown leadership in the grid-application world, and application adoption is what will drive grid," says Platform's Baird, adding that grid is most appealing to vertical industries with "heavy computational work," including finance, manufacturing and life sciences.

Still, he acknowledges that it would be difficult for just any VAR to begin a widespread deployment of grid offerings from scratch. Even the ever-hyperbolic Ellison acknowledges that grid has basically no chance to be an overnight sensation.

"People hate change, and we know we're making this presentation for the future," he says. "The large mainframes and other machines won't disappear overnight, but this begins the gradual move to a grid approach, and it's an inexorable move from where we are today."

'Rip And Replace'

Fortunately, the grid model lends itself nicely to a gradual adoption path.

"The cool thing is you don't have to rip and replace everything," says Benny Souder, Oracle's vice president of distributed database development. "You can start small with a few devices like blade servers and incrementally scale out from there."

However, even if grid makes a quicker impact on life sciences and other research-oriented arenas, the tougher sell will be convincing everyday companies to almost totally rework their networks over time. With that in mind, Oracle is training its staff to sell 10g. The company is also spearheading efforts to create a consortium of grid vendors to establish a set of business-specific grid standards.

One potentially huge edge for Oracle's version of grid over its competitors could be cost. For example, at OracleWorld, Ellison said his version of grid primarily uses two-processor, fault-tolerant machines instead of larger, more expensive servers, enabling Oracle to offer a staggering cost difference of roughly $860 per GHz; IBM's offering costs $27,000 per GHz.

"At the midmarket level, a grid made up of two-processor machines is faster and more reliable than four-processor servers," he said. "Resellers will love this because the initial outlay is very manageable, and it provides a graceful upgrade path."

IBM accuses Ellison of voodoo economics. Regardless of who's right, there's little doubt that a fully implemented grid framework will save money over the long haul. It's just a matter of timing.