[Pc_Support] Ziff-Davis Sensationalist Headline: "Study Finds Windows More Reliable than Linux"

[Bryan J. Smith <b.j.smith at ieee.org>]

I've read through Microsoft's new study entitled
"Microsoft Windows Server 2003 vs. Red Hat Enterprise Linux AS 3.0:
IT Professionals Running a Production Environment (04/05)"

Which is located here:  
http://www.veritest.com/clients/reports/microsoft/microsoft_IT_Pro.pdf  

The study _induces_ several scenarios and tracks the response time of several carefully qualified individuals on both Windows and Linux.  All-in-all, GIVEN THE PARAMETERS (this is a key note ;-), the study does not look biased at all to me.

"Key findings
  - In our tests, the Windows Server 2003 environment had
4:20:19 of average end-user service loss time compared to
4:59:44 of average service loss time for the Red Hat
Enterprise Linux AS 3.0 environment on measured service
loss events. Lower results are better.
 - In our tests, more work was completed in the Windows
Server 2003 environment (280 completed tasks and events)
than in the Red Hat Enterprise Linux AS 3.0 environment
(248 completed tasks and events) in less average elapsed
time (18:44:14 vs. 27:48:05)."

Linux (and UNIX) often takes me a good 25% longer to resolve issues than Windows.  Of course, I _never_ get unresolved problems or blind reboot fixes.  And in many cases, I never see an issue until I reboot -- which is always rare and scheduled off-hours (whereas Windows has caused me grief at the worst times).

I have documented why this is so -- largely the "piecemeal nature" of UNIX/Linux services makes it more involved to configure or resolved, but it's very, very reliable and rarely needs to be touched.  Also, the lack of reboots means you rarely have to deal with reboot-time recovery, and when you do, it's entirely off-hours (never had an issue to the contrary except once, long ago).  NT is the exact opposite, quick to configure/fix, but lots of random occurances, chance induces (largely due to integration of "Chicago" code and other details I've detailed before), and the constant requirements to reboot.

But given the parameters, that _both_ systems would have the _same_ induced issues, I have no issue with the study and report, it was well-done from that standpoint.  It does take me less time to resolve issues on NT than Linux.

So what's the uproar?

Check out the Ziff-Davis headline, the absolute *WORST* PUBLISHER FOR SENSATIONALIST TITLES THAT ARE *0* REPRESENTATIVE (written to tell CIOs who don't read them 180 degrees opposite of the actual article content):  
http://www.eweek.com/article2/0,1759,1782996,00.asp

[ Omission:  I write for another major IT publisher, and I have been grossly misquoted by Ziff-Davis in the past, so I clearly have a bias. ]

THIS WAS *NOT* A TEST OF RELIABILITY!
I don't doubt this was an editoral move, and not the author's fault.
I've seen good writers for Ziff-Davis have their titles totally gone 

It was a test of "recovery" when things do go wrong.
And that assumed Linux would suffer similar mishaps to Windows.
Again, given the parameters, it was a fair test.

But the parameters were *NOT* REAL-WORLD!
This fact has been highlighted in a recent admission by the vendor who did Microsoft's previous TCO study.
They have found that their own surveys of real companies show that Windows is not cheaper than Linux from a TCO standpoint.
Because, as I posted before, the parameters of that study did not match "real world."

But even in that case, Ziff-Davis did not got his far.
I am sick and tired of Ziff-Davis "stiring up XXit" just to make headlines.
Don't blame Microsoft for this one, the VeriTest title was accurate.

But Ziff-Davis continues to be one sensationalist pig of a media outlet.



--
Bryan J. Smith   mailto:b.j.smith at ieee.org