Posted by: Steve Hamm on May 6, 2005
It’s amazing how a an architectural shift in one technology can upset the apple cart for another. That’s just what’s happening as a result of the shift to produce microprocessor chips with more than one processor on them. Since many software companies charge their customers for products based on the number of processors in a server computer, this is making them rethink the way they charge. It’s also making some customers squawk. Interestingly, this tempest may result in the tech world shifting over to utility computing—customers paying for processing power only when they need it—even faster than pundits expected.
This matters a lot. The sooner tech companies come up with smart ways to deliver computing on demand, the better off they and their customers will be. It's all about making the complex parts of computers invisible to the people who are using them.
The processor condundrum started to emerge four years ago when IBM introduced the first dual-core processor. That means two data processing devices are mounted on a single chip. It's all about increasing the power of a server without making the machines so hot that they literally catch fire. Later, Sun followed. And now Intel and AMD are delivering dual-core chips for servers. Soon there will be multi-core chips.
When the shift started, the software companies kept charging per-processor. But last October, Microsoft broke ranks. The arrival of dual-core processors from Intel and AMD, often used to run Microsoft's Windows operating system, forced a re-think. Microsoft decided to treat dual-core processors as if they are just one processor. That lit a fire under Oracle and IBM. So far, Oracle is sticking with its old policy. But IBM just a couple of weeks ago came out with a hybrid approach. It still charges per-processor for software running on its powerful Unix servers, but charges per-chip, like Microsoft, for servers running Intel and AMD chips. Both Oracle and IBM also alternative ways of charging, as well. Here's AMD's recommendation for how software ought to be licensed.
Software companies have a whole bunch of ways they charge for server software. They include site licenses, fees based on the number of computers, fees based on the number of users, and others. Eventually, according to Margaret Lewis, senior software strategist for AMD, utility computing will be the new model for charging. "Customers are going to demand a pay-for-what-you-get model. The faster the software vendors figure that out the better off they'll be," she says.
Of course, it's easy for a chip maker to lecture software makers on what they ought to do. But the software giants are heading in this direction. The only question is how long it takes them to get to the destination.