The Great Grid Partner Grab

What if there were a technology that offered a legitimate excuse to advise customers to expand their networks with additional servers and workstations, as well as the chance to sell them the software to exploit all that idle computing horsepower? You'd jump at it, wouldn't you?

Well, for a broad assortment of integrators, ISVs and VARs, grid computing is just that technology. Less than a decade ago, grid was the domain of geeks working with Cray Research supercomputers in glass-enclosed data centers. Today, it's the buzzword du jour, embraced by IBM as the centerpiece of its on-demand strategy and positioned by Sun Microsystems as the hot technology that will take the company's vision of network computing to the next level.

In many ways, grid is perhaps the biggest new opportunity to come along in the channel in years, as grid players anxiously look to line up partners. Specifically, three high-profile Tier 1 vendors—Hewlett-Packard, IBM and Sun—are all making grid moves and seeking partners. And an oligopoly of bigfoot ISVs, which dominate the grid space—United Devices, Platform Computing and DataSynapse—are also seeking VARs to help stoke sales.

Clearly, the market opportunity is big. According to Insight Research, Boonton, N.J., worldwide grid spending will grow from an estimated $250 million in 2003 to $4.89 billion in 2008. For VARs, selling even a small grid can mean a big payday: A setup with 50 to 100 CPUs will run about $50,000 (depending on the services and hardware bundled in), and larger grids can go into the millions.

But if grid sounds like a reseller nirvana that's poised to take off, it is, nevertheless, fraught with pitfalls for the unwary VAR. For one, there's a steep learning curve, which makes it much more complex than simply selling and integrating out-of-the-box hardware. While no company has done more to build the cachet of grid than IBM, HP and Sun also have heated up their grid offerings in recent months. Lately, grid vendors of all stripes have been spewing forth heavy marketing hype, sowing some customer confusion in the process.

"Nobody can seem to really agree on what grid is," says Ian Baird, CTO for grid and utility computing at EMC in Hopkinton, Mass. In reality, grid has several facets, Baird says. "It's distributed computing on steroids," he explains. "It's typically heterogeneous, so you're connecting together all sorts of different compute resources from different vendors."

For solution providers wondering whether to get in on the ground floor, the unanswered question remains whether grid computing is overhyped. Experts agree that it pretty much depends on what you call grid. If your definition centers on the lower-end technologies, such as utility computing (see "Utility And On-Demand Differ From True Grid," below), then grid probably has promised more than it has delivered. If, however, you define grid as a "true grid" of interconnected clusters, then maybe it's just coming into its own.

"Things like utility computing and on-demand are business models for delivering computing," explains Greg Papadopoulos, CTO at Sun Microsystems. Papadopoulos has perhaps a more rarified definition of grid technology. "Grid is an example of network-scaled computing," he says.

Lest you think that's an overly scientific definition—Papadopoulos used to be a computer-science professor at MIT—he readily relates it to a real-world example. "Network-scaled computing is something like Google, where you've got hundreds of thousands of processors that are being applied to a task," he explains. "The only way you can think about doing this is by placing the network at the center of all those machines."

In such wildly complex environments, the challenge for grid providers, Papadopoulos says, "is, how do you program them, what is going to be the operating environment, how do you manage them, how do you do [on the network] all the things that you'd do on a single system?"

Those are precisely the questions asked by vendors and VARs who've wondered how they can prove to customers that grid is a technology with a quick payoff. The answer seems to be, "Start small and get your foot in the door."

"We've sold some very large hardware deals where the initial thought was, 'All we're going to be doing is some services and nothing else,'" says Steve Gordon, the executive in charge of IBM's grid-computing organization. He points to brokerage house Charles Schwab, which in 2003 enlisted Big Blue to install a pilot grid that's currently undergoing a large-scale expansion.

"What was initially sold was an in-house application they needed to run while they were talking to a customer," Gordon explains. "High-wealth clients would call with a question about some financial calculation. It took too long to get an answer, so [Schwab] had to hang up and call the customer back. Now, with the grid, they get the answer right when the customer's on the phone because the application runs during that time frame."

When the pilot program proved a success, cutting some customer wait times from 4 minutes to 15 seconds, Schwab signed on IBM to expand the grid.

Such a two-step approach to rolling out a grid is a typical strategy. "Most deployments start small," explains Ed Hubbard, president and founder of grid ISV United Devices in Austin, Texas. "Everybody has heard about grid computing, but most people don't really know what it can do for them. We usually do a pilot, where we take an application and migrate it onto a grid made up of Windows and Linux servers, cluster nodes, nondedicated devices and workstations."

The network is usually up and running—with the grid application installed and buzzed out—in a day or two. "Then you can show the customer what kind of return they can get on their grid," Hubbard says.