The lingo, complexity and sheer scale of on-demand computing often seem so vast as to be paralyzing. But when you peel away the marketing spin and the many terms used to describe this phenomenon, you can reduce it to two phrases that everybody understands: "waste not, want not" and "pay as you go".
While the concept may be simple, the execution isn't. Full on-demand computing at the enterprise level requires degrees of reporting and management unheard of in most companies. It also entails rethinking the organizational structure of IT -- focusing on business needs rather than on technical specialties.
In an effort to hack through the definition thicket, here's a simple way to think of on-demand computing: It's when computing resources are made available to users on an as-needed, pay-per-use basis. These dynamically allocated resources may be maintained within the enterprise itself, or they may be procured through a service provider.
Many in IT use the term utility computing as a synonym for on-demand computing, but utility implies near-total use of outsourcing, application service providers and managed services. For some business executives, the idea of turning on a tap and getting all the IT you need -- along with an easy-to-read bill at the end of the month -- is appealing. But it will be a long time before this is the reality in any sizable enterprise. Existing investments in infrastructure, as well as the desire to use IT as a competitive differentiator in selected strategic areas, make the 100 percent-utility model unfeasible for most.
However you define it, on-demand is in demand. In a recent Computerworld online survey, 32 percent of the 765 respondents said they are pursuing, piloting or have deployed on-demand computing, while 30 percent are investigating it. Economics is the primary driver
Emphasis on cutting costs
In 2003, when health care insure Cigna began talking with IBM about on-demand, cost reduction was the principal motivator, according to Ben Flock, vice president of virtualization. Cigna was using more than 3000 servers for myriad applications, including self-service portals used by many of its 10 million customers. "We'd been developing apps on fly, so we had a lot of legacy [systems]," Flock says.
Initially, Cigna's goal was simply to move all applications to a single version of IBM's WebSphere "to drive down operating expenses," he says. But when IBM said Cigna was a candidate for on-demand, Flock agreed.
Other on-demand pioneers also turned to the computing model to save money. Two years ago, when Exxon Mobil in Texas, made its Mobil Travel Guide unit an autonomous division, its CIO Paul Mercurio went the service-provider route, opting for IBM's On-demand Linux Virtual Services to host and manage the division's e-business applications. "Despite the Exxon Mobil name, we were acting as a start-up. We knew we would grow rapidly, and our business was seasonal," he says.
Every major enterprise software vendor offers on-demand in some form, depending on that vendor's history and strengths. Some focus on management and reporting tools, others on virtualization and still others on the services model.
But a successful on-demand strategy demands organizational change as well. CIO James Dallas is trying to retool the IT mind-set before introducing on-demand products and services. His goal is to help his 1200 IT employees think more like business people before his conglomerate shifts to on-demand.
"We need to run IT like a business, but what does that really mean?" asks Dallas. "To someone who's been in IT all his career, that's an empty phrase."
To make the phrase meaningful, the company is putting all its IT employees through financial-skills workshops and other fiscal training. Dallas has also tasked a communications specialist to constantly communicate the changes the new model will bring, and the company has formed centres of excellence around three key areas -- business intelligence, server support and e-commerce -- in which on-demand will be put to the test. "We're still early in the process," Dallas says, "but we think it's important to adopt on-demand as a business model, then put the technology in place."
Alignment is everything
If there's an under explored aspect of on-demand, it's that the model can't succeed in one of its primary goals -- aligning IT use with business needs -- unless coupled with a business process methodology. Some of the methodologies designed to do just that are IT Information Library (ITIL), balanced scorecard, business service management and portfolio management.
These disciplines force organizations to begin at square one, which may prove frustrating for IT managers expecting quick benefits. Even basic nomenclature must be addressed; if storage capacity or administrator isn't defined exactly the same way across all of an enterprise's data centres and IT groups, it will be impossible to pool and manage these resources. According to Chris Buss, manager of virtualization at Hewlett-Packard, this lesson was learned by Philips Semiconductors when the manufacturer implemented the Utility Data Center, HP's flavour of on-demand.
"Philips quickly found out it needed standard nomenclature in place," Buss says. "So with help from us, it used ITIL to create a set of universal definitions they called the Soldier's Handbook." Buss says Philips has since achieved 45 percent cost reductions in its data centre operations.
Early users notwithstanding, on-demand has until recently been shrouded in unrealistic expectations and conflicting definitions. Fortunately, the peak hype has passed, replaced by sober assessment of the model's limitations. "There are a number of business objectives on-demand can help solve," says Bill Mooz, senior director of utility computing at Sun Microsystems. But it won't magically slash IT expenses on its own, he adds. "If your cost is too high, it may be because you're using the wrong architecture or operational model," he says. "You need to address the root cause before you start worrying about [on-demand] as a financial model."
What needs to be done to push on-demand forward? Analysts say apples-to-apples pricing comparisons would be a major step. "All the big systems vendors have noted this confluence of (distributed enterprise computing) trends and have responded with a blend of services and highly distributed software," says Dan Kusnetzky, an analyst at IDC. "But they've all spun it their own way; there are now a dozen buzzwords" for on-demand.
Kusnetzky says on-demand will find greater enterprise acceptance when terminology and pricing models grow more standardized. Flock says Cigna's move to an on-demand model for its IBM WebSphere operations has reduced costs significantly. Moreover, he adds, the maturation of the discipline promises to bring about the hoped-for tie-in between business needs and IT resource allocation.
"The initial drivers for on-demand were reduced operating expenses," Flock says. "But we're seeing other benefits now, like improved security and performance guarantees. This looks like the real thing."
Terms to know
Utility computing, sometimes called metered services, is a subset of on-demand computing in which IT resources are provided by an outside services firm, and the client company pays only for what it uses. (To many, utility computing and on-demand computing mean the same thing.) Grid computing is the practice of applying multiple computers' processing power to a single problem simultaneously. Created as a way to use computers that would otherwise sit idle to solve complex scientific problems, grid computing is still most frequently used in the technical, scientific and academic communities.
The virtual data centre entails pooling resources to make individual components behave as if they were a single entity.
Autonomic computing describes an infrastructure that monitors itself, allocating resources and even making repairs without human intervention. Task list
An enterprise on-demand strategy demands that companies do the following:
- Make existing IT resources (including personnel) more flexible, so they can be used by business lines according to need and priority.
- Improve reporting and monitoring tools. This will help IT ensure that the right resources flow to the right places at the right time and that departments and lines of business are charged appropriately for their usage.
- Determine which technology needs are commodities that can be outsourced or purchased on a pay-per-use basis (storage, perhaps, or the help desk) and which are strategic and should be kept in-house (such as a highly developed CRM system that's unique in its industry).