No Docs?

    June 7, 2006

Microsoft can buy white papers whenever it wants, and of course they do. They say some of these are “independent, non-sponsored” studies.

Yeah, right. If you look closely, you’ll find the same companies that do “independent” studies also have done “Microsoft-sponsored research”. Sure, I trust their “independent” stuff without question.

Anyway, apparently the most recent Yankee Group study has found that Windows servers are more reliable than Linux, but that Unix servers are more reliable than either. I can’t find anything on Microsoft’s website referencing this (I guess that’s not surprising) though they do have plenty of other studies.

The Enquirer’s review of the Yankee Group study says:

Yankee says that Windows 2003 Server is a more reliable server operating system than Linux as the downtime was longer for Red Hat Enterprise Linux, and Linux distributions from “niche” open source vendors. It says that one reason for this was that there was a shortage of Linux and open source documentation.

Huh? A “shortage of Linux and open source documentation”??

I’m still shaking my head over that one.

PC World covered a similar study. They report:

In the absence of a clear definition of reliability or benchmarks, Microsoft commissioned a study that pitted Windows Server 2003 against Red Hat Enterprise Linux 3.0 AS. As part of the study, 18 Linux and 18 Windows system administrators were hired to run the simulated IT environments of a midsized company over a four day period.

That’s not the Yankee Group study. This is one Microsoft paid for and can be found at Microsoft Windows Server 2003 vs. Red Hat Enterprise Linux AS 3.0: IT Professionals Running a Production Environment

We’ll assume that there was enough integrity that the Windows administrators weren’t ten year veterans pitted again Linux noobies fresh out of community college. There was testing of the admins hired for this study. However, the testing wasn’t much for either the Windows or the Linux folk. The Linux test questions were slanted toward Linux in a Windows environment, which is not totally unreasonable but definitely not always the case. The Windows test didn’t take any notice of a Unix or Linux environment being present though, and in fact that is pretty likely at larger organizations and even quite a few small shops. Overall, the qualification questions were fairly basic.

PC World continues:

The administrators had to troubleshoot intentionally introduced errors and conduct tasks such as configuring new devices, making backups and setting up remote access.

What sorts of errors? Were the “errors” truly similar? Were they equally likely to happen?

You can download the PDF and read it for yourself. My take is that it was a bit slanted toward Microsoft environments (heavy concentration on user file deletion and SMB browsing issues), but in general no competent Linux admin should have had any difficulty reacting to the tasks and induced problems and solving them quickly. In fact, the results seem to bear that out: there were some Linux and some Windows admins that got stuck in certain areas, but most solved the problems in the time frames I’d expect.

As hinted at above, it might have been more fair had the Windows folks had to interface with Linux or Unix machines, but they didn’t have to. Linux admins did have to deal with Samba and Active Directory issues.

The study ignored the sorts of tasks that are common in any environment. For example, let’s ask the admins to create a file mapping IP addresses to host names, using a naming format where (for example) is host_2_1. Almost all Linux/Unix admins would script that and be done in seconds, while a depressing percentage of Windows admins would fire up Microsoft Word and start typing. Of course that’s really nothing to do with the OS proper, but it does show WHY Windows servers usually do require more work: their admins lack both proper tools and the skill to use them.

Overall I’d say this whole study is meaningless. There’s tremendous variance between the various admins and variance between similar tasks for the same admin. This is simple reality for troubleshooting: sometimes even the best of us head down the wrong path for a while. I don’t find the time differences compelling.

For me, it doesn’t change the basic observations every Unix/Linux tech knows: once the system is properly set up, it will probably run for years without attention, and any organization with Windows servers or desktops will always have a larger number of support techs than the equivalent Unix/Linux shop. Those are the real facts, and Windows compares disfavorably.

*Originally published at

Add to | DiggThis | Yahoo! My Web | Furl

A.P. Lawrence provides SCO Unix and Linux consulting services