% fortune -ae paul murphy

Updating perceptions

My blogging collague Dana Blankenship recently committed a horrible anachronism with the comment that:

It would be very bad if Google were stuck with Sun's overheads, running expensive SPARC servers instead of cheaper Linux boxes.

There are two things wrong with this: first, Solaris on SPARC is actually more than competitive with Lintel and, secondly, Sun now makes the fastest and least expensive Lintel servers around.

Dana's mistake is ridiculously common - people don't update existing certainties until forced to by something external. You can bet, for example, that Dana now knows about Sun's Dell beating opterons -and the fact that Solaris for x86 really is free-er than Linux.

On the other hand neither of us yet knows which would really be cheaper for a company like Google - running on SPARC/CMT or running on Opteron. As I mentioned on Monday, however, we may be about to find out that Solaris on SPARC/CMT, not Linux on Opteron, wins this cost/performance contest.

I stumbled across a much more extreme example of anachronistic opinion recently while talking to an internet and open source luminary who questioned my perception that Solaris is significantly better (on grounds of flexibility, reliability, and organizational performance) than Linux. His understanding of Solaris, it turned out, is based on personal experience - with SunOS 4.1.4 on something prior to the SPARC 10 in the early ninties!

That may be extreme - but walk down the hall wherever you work and ask people which costs more: Dell or Sun and almost everyone will give you the wrong answer. Ask almost any senior decision maker in Finance what a Unix server costs - and you'll get numbers in the half million and up range. Ask the same person what a PC costs and you'll usually get all-in estimates below the price of the Office software he'll tell you the machine comes with.

In other words, one perception was set in intellectual concrete years ago, the other by some pharmacy's adsert in this morning's newspaper.

Like, the world changes, eh? But you know what's truly frightening? that's how the majority of IT decisions get made: on unexamined, usually anachronistic, certainties that get passed along as fact between individuals and through the media.

Computerweekly, for example, carried a story by Lindsay Clark last week under the heading: University's Linux migration cuts costs and boosts SAP performance that described Stirling University's migration from HP-UX on PA-RISC to Linux on Itanium in glowing terms:

"The Integity systems offered better price/performance than comparable Risc-based systems, with the added bonus that we will maintain flexibility in choosing any operating system in the future," said Martyn Peggie, HR IS manager at Stirling University.

We can choose HP-UX, Red Hat Enterprise Linux or Windows. This will allow us to adapt to changing requirements over the years to come," he added. "We believe that with the current configuration of the IT systems, the servers will pay for themselves over the next three years just by savings in support costs."

The transition, as both this report and the source document proffered by Red Hatgo to great pains to point out, gave these geniuses a whopping three times performance improvement over those archaic and high cost RISC boxes - which have been running this and other applications since the April, 2000 go-live date for the SAP components.

So if you're looking at a hardware or software decision any time soon - start by listing the things you know for sure about relevant pricing and then go find out if any of them are true before committing further.

Paul Murphy wrote and published The Unix Guide to Defenestration. Murphy is a 25-year veteran of the I.T. consulting industry, specializing in Unix and Unix-related management issues.