Sun Microsystems Inc. yesterday unveiled its new on-demand pricing model, but some CIOs said they aren't ready
to buy the promise of cheap computing capabilities for IT operations.
The Sun Grid will cost clients $1 an hour for each microprocessor used and $1 per month for each gigabyte of storage. The announcement is the latest in a series of moves by Sun to simplify its pricing models.
The idea is that users can obtain the extra computing power they need, when they need it and can link into Sun's grid network.
But IT executives who are considering on-demand computing models must balance the low-cost offerings against their overall business needs and concerns, such as security, according to analysts and CIOs who reacted to Sun's announcement.
"At the end of the day, you really have to look at your business model and whether that works or not," said Marty Colburn, the chief technology officer at the National Association of Securities Dealers (NASD) in Rockville, Md.
"It's intriguing, but it's back to the future. It's TSO [time-sharing option] all over again," Colburn said.
For Colburn and the NASD, where transactional processing is key, the Sun Grid model probably is not a good fit -- because of security concerns in the transaction heavy business, Colburn said.
"I think it could work, but I think it remains to be seen what kind of reaction they're going to get at this point in time," Colburn said.
Harry Roberts, CIO at Boscov's Department Stores LLC in Reading, Pa., said his company doesn't currently use Sun hardware, but "if the on-demand type of model could be shown to actually save money, it is worth a look."
"We have been working with IBM to review the potential of on-demand computing for Boscov's, but we are still doing some fact-finding," Roberts said.
"In a nutshell, we would be interested in any approach to CPU pricing that saved us money and provided flexibility over our current model," he added.
Presently, Boscov's runs IBM hardware in its data center, including IBM zSeries mainframes, for its retail store IT operations. Roberts said the option to "regulate" grid computing purchases is very important to him. "I don't want a surprise at the end of the month" when the bill arrives, he said.
Chuck Kramer, CTO at Social & Scientific Systems in Silver Spring, Md., said that grid computing wouldn't work for his company because it deals with Medicare data that contains confidential personal information -- data that he wouldn't want to send through another network.
Still, Kramer said the grid pricing model might make sense for disaster-recovery proceses. "The downside is that you don't know how much it will actually cost you until you run your jobs. You can run your costs up really fast."
Rikki Kirzner, an analyst with Waltham, Mass.-based Hurwitz & Associates, said Sun's new on-demand strategy has potential for users, but that getting applications and data to the grid could be tricky. "There's opportunity, but a lot of heavy lifting between there and here," she said.
"It's early, but it's an interesting take," said Kirzner, who attended the Sun briefing at the company's Santa Clara, Calif., offices yesterday. "They have all this horsepower and they've been talking about the network being the computer for God knows how many years."
Sun is likely hoping that if IT users aren't ready to buy more hardware, perhaps they will buy more computing power, she said. "Ultimately, their goal is to make it a vertical market play."
For traditional retail businesses, manufacturers and health care industries, there's a long way to go before Sun Grid could work for users, Kirzner said. However, it is a viable option for industries that demand incredibly high computing power, such as animation filmmaking, oil and gas exploration, and the pharmaceutical and biotechnology industries, she said. "This could provide power they don't need to own," she said.
"There's a lot to be done between the dream and the reality," she added.
The Sun Grid will operate using Sun's Sparc and Sun x86 platforms and software, running Solaris 10 and the Sun N1 Grid Engine.