|
Utility Computing: A Loss Leader for Vendors? I attended a number of sessions on the topics of utility/grid/autonomic computing at both COMDEX and the Jupitermedia cdXpo conference last week. I've been studying these topics recently, and I'm having trouble accepting one of the supporting arguments: that systems are too expensive and that we therefore need to increase our utilization ratios.
It's deja vu. In the 1960s, mainframes were so expensive that we had to keep them running 24x7. Staff was cheap, but the hardware cost a fortune, so we queued up jobs to run at all hours of the day in order to keep the CPUs busy. Then, in the 1970s, came the minicomputers--the first computers we could afford to turn off at night. Also, the first computers we could generally afford to dedicate to single tasks, even if they weren't full time. The PCs of the 1980s continued that trend: hardware and software that was so inexpensive, we could dedicate it to the individual.
So what's changed in the 2000s? Hardware is cheaper than ever, but for the past three years, we've been in a recession, and IT managers have been asked to do more for less. I think utility computing *may* be a vendor-centric answer to the spend-less demand. The vendors don't really want to hear "spend less." Utility computing is their way of re-wording the requirement as "spend more for more." But that's not the real requirement. They're not listening.
I think there's a good chance the overhead and complexity of utility computing will outweigh its benefits. That's why SANs cost so much more than the underlying disk drives, and server clusters do likewise. Discrete systems are always less expensive than linking them together to squeeze every last drop of performance out of them.
Once the economy recovers and IT budgets rise, the demand for spend-less will diminsh, as will--I predict--at least this argument in support of utility computing. What we'll continue to see is the ever-increasing cost of software as a percentage of overall system cost. I expect this to be the greatest area of IT manager dissatisfaction, and the driving force behind the increasing use of open-source solutions. Oracle (in particlar) and Microsoft will have to respond to the need for lower pricing if they're to beat off the open-source threat.
I listened to a few case studies of utility-computing success, and I'm very suspicious of their economics. The customers claim lower costs than using traditional solutions. But if you dig deeper, I think you may well find that all of the systems in these early case studies have been sold as loss leaders in order to generate buzz surrounding the utility-computing concept. When you price these systems out using legitimate street prices, I think you'll find they're very expensive.
Posted Tuesday, November 25, 2003 2:12:42 PM 
|
