Since the dawn of the Industrial Revolution, a steady chorus of Luddites, doomsayers, and social critics has voiced concerns about the impact of technology on mankind. Usually, this concern has come in the form of a simple question: At what price progress?
At the end of this year, finance chiefs will have a pretty good idea of the price. That’s when IBM is scheduled to change the way it charges customers for use of the company’s middleware products. Managers at IBM say the new pricing model (which affects about 350 products and 1,300 part numbers) was necessitated by a number of trends buffeting the IT industry. But the reality is, IBM is simply bowing to progress — in this case, progress in server technology.
Indeed, recent advances like virtualization, grid computing, and, most importantly, multicore processing, have radically altered the old concept of per-processor licensing models. “It’s hard to determine which application is running on what processor,” acknowledges Jeff Tieszen, a spokesman for IBM. “Pricing based on processors is outmoded.”
Rather than try, IBM will now charge clients based on the performance of underlying hardware — what Big Blue dubs “processor value units” (PVUs). In the past, IBM has followed the industry practice of basing fees on more-general measures, like processors or cores, which are the computational engine of a microchip. But given the variety of chip designs out there, and with more on the way, many analysts think the switch to processing output makes sense. They also believe other software vendors will take a similar tack. Says Amy Konary, a program director at technology research firm IDC: “The days of licensing software on a simple one-license-per-installation basis, or on a straight per-user or per-processor basis, are coming to an end.”
In theory, performance-based pricing should help IBM better match licensing fees to the actual mileage a customer gets from a program. The company has been basing fees for business products like Lotus Groupware and Websphere middleware on the number of cores in a processor. For an x86 chip with two cores, IBM has been charging half a licensing fee per core. For an existing RISC-based chip, the vendor has been charging one license per core.
Adding to the muddle: different vendors use different methods for pricing applications that run on the same chip. This not only complicates things, it tends to make people’s heads hurt. “It was starting to get to the point where customers were having a hard time budgeting for next year,” says Tieszen. “They didn’t know how new processors would be priced.”
Hence, IBM will base its pricing on PVUs — essentially, the processing muscle of a chip. Analysts say that’s important, since some single chips with dual cores do not produce twice the processing power, as you might expect they would. Instead, a number of dual-core chips generate only one-and-a-half times the processing speed of a single-core chip. “In that case,” asks Dwight Davis, vice president of market-research firm Ovum Summit, “why should you pay the software vendor for two full licenses?”
Likewise, clients often end up paying more if their software runs on a RISC-based computer. Why? Because a RISC machine typically contains less-robust processors than an IBM server, and thus needs more of them. More processors means more licenses. In addition, users frequently pay the same per-chip fee for an application that’s running on a dual-core processor, even though one chip may perform twice as fast as the other. “There hasn’t been any set model, so vendors have been pricing their software a number of different ways,” says Tieszen. “We think [performance-based pricing] is at least a step in the right direction.”
Others aren’t so sure. Critics contend that performance-geared pricing will be just as confusing as other approaches, requiring customers to pull out slide rules and consult complex charts. A few worry that IBM may be tempted to lower the performance ratings on its own chips, thus making the company’s enterprise applications more cost competitive.
So far, a number of software vendors have announced they will not be following IBM’s lead. Notably, managers at the world’s largest application publisher, Microsoft, have indicated that the company won’t be adopting performance-based pricing. Instead, Microsoft will continue to charge a per-processor licensing fee for programs like Excel and Word. In a published interview, one Microsoft executive ripped performance-based pricing, insisting it will jack up the fees on software.
He may have a point. Certainly, it’s not likely that IBM will be giving away the store with its new licensing approach. As IDC’s Konary notes, the technology giant is making a strategic bet with the policy. And while she doesn’t expect any short-term impact on IBM’s software prices, Konary concedes that the company’s licensing fees could go up as processor performance increases.
That sort of performance increase is a given, with chip-maker Intel already planning to launch quad-core chips in the coming year. “As with any licensing change,” Konary notes, “only time will tell whether customers will view [IBM's] new approach as fair and equitable.”
One possible indicator: Oracle unveiled a similar scheme (based on “universal power units”) back in 2000. Within a year, the database leader had scrapped the controversial plan amid a flurry of complaints that it penalized clients who — wait for it — used more-powerful servers.
In all likelihood, IBM’s new software-pricing plan will probably not be the final word on licensing fees for multicore and virtualized environments. Davis says a growing number of companies are moving toward flexible data centers. Under that setup, network configurations can be quickly shifted, with software that’s running on one system swapped onto another.
Experts say such an approach puts a premium on scalability and flexibility. Licensing schemes will no doubt need to reflect those attributes. “Ultimately, people will want licenses that allow them the flexibility to scale up and down and move to different systems,” says Davis.
Overwhelmed? Join the club. A bit of good news, though: IBM’s new licensing policy will have little impact on software running on existing machines. For those applications, Big Blue is merely changing the terminology it uses — from “per processor licenses” to “processor value units” — and multiplying the current license requirement by 100. When new processors enter the market, IBM will assign them a number of PVUs based on relative performance. Tieszen says the company is purposely taking it slow with its new approach. “We want to get customers to the point where they understand the new model and adjust to it.”
They’ve got two months to catch on. The new pricing model will take effect with Intel’s new Clovertown quad-core chip, scheduled to ship in December.
Esther Shein covers technology from Framingham, Massachusetts.
What’s under the Hood?
With IBM’s new pricing plan, processors will be given a power rating.
|Single core (all platforms)||100|
|Power 5 QCM dual-core||50|
|Sun T1 octi-core*||30|
|* T1 entitlements per processor adjusted with the elimination of fractional licenses.