Todays Weather Report

Talk about remote control: When Eli Lilly’s Dave Powers needs to fire up a rack of servers so scientists can run an experiment modeling, say, a potential blockbuster drug, he pulls out his iPhone and taps the screen. When the job is done, he flicks his finger and the computers disappear. Literally.

That’s because they don’t really exist—at least not on the premises of the pharmaceutical giant. They’re located in the “cloud,” or to be more exact, at one of Amazon.com’s massive data centers. Lilly rents computational time from the virtual vendor, paying only for the firepower it needs at any given moment. When a job is done, the meter stops ticking.

“While we’re talking I can take out my iPhone and literally spin up a 100-node Linux cluster and do some genomic analysis and it’ll cost $10,” says Powers, an associate information consultant for Eli Lilly’s research and development unit. “When we’re done, we tear it down. We don’t own it and we don’t pay for the costs of storage, the data center, or the people to support it.”

In the old days—nine months ago—it would have taken seven to eight weeks for Powers to deploy enough servers in a Lilly data center so scientists could conduct that experiment; today, it takes five minutes to get them online at Amazon. “I can go to a scientist and say ‘I can give you eight weeks of time for your research, what’s that worth to you?” notes Powers.

Startups and small businesses were the first to embrace cloud computing, in some cases offloading their entire I.T. operations. But Fortune 500 companies are also starting to look skyward. “In terms of trends, a lot of people are doing what I call science experiments, investigating using the cloud informally or for large-scale computer jobs that run infrequently,” says Joseph Tobolski, director of infrastructure R&D for Accenture, the global management consulting and technology-services company.

Driving the move online is the bleak economy and budget-slashing bean counters. “Every CIO I speak with is looking to avoid capital expenditure and decrease risk,” says Adam Selipsky, Amazon Web Services’ vice president of product management and developer relations. “They’re trying figure out how to do the same or more with less.”

Tobolski estimates the savings from cloud computing can be substantial. A recent Accenture report that he co-authored cited the example of a global logistics company that could spend $4 million to buy 150 servers plus $1 million in annual software license fees. Contrast that with the Amazon cloud services fee of $131,000 a year for everything—a price not much more than the $70,000 a year that would be needed just to keep the in house servers powered up.

Powers has yet to calculate the dollars-and-cents savings that came from moving some of Lilly's R&D to the cloud. But he says that leaving the company’s in-house data centers behind has already unleashed his scientists' creativity. Now Powers can launch unique computing environments on the fly for scientists’ individual projects. That allows them to quickly test hunches that in the in-house computing era would have gone unexplored.

When it comes to cloud computing, the only problem, says Tobolski, is that “the hype level is at DEFCON 2” and the whole concept needs a reality check. No large company is going to scrap all its data centers and beam its entire IT operation into the clouds. Rather, the future, he says, probably will belong to the hybrids—those companies that combine a private in-house cloud for mission-critical operations while using the public cloud for more routine back-office work and heavy-duty computing like Lilly’s pharmaceutical research.

The key, according to Tobolski, is to carefully evaluate which jobs are most suited for cloud operations and in what order they should be moved out of the in-house data center. Security, privacy and the reliability of the cloud continue to be concerns as well as worries about tying one’s corporate fortunes to a cloud provider’s proprietary format. Powers concurs, noting that Lilly didn't just dump existing IT operations on remote data centers; rather they took the time to redesign internal processes to best exploit the cloud’s computing power. “The cloud is not the cure for cancer," says Tobolski, "but there are tremendous opportunities out there.”