This article talks about a campaign Microsoft has apparently started to deliver a few “facts” about the reliability of Windows vs that of Linux. It’s almost certain to generate a huge flamefest with little useful dialog for those of us that need to weed through all the rhetoric and make good choices for our companies or clients. It did remind me of a point though that for some reason I don’t see made very often out there.
There is a common view, even among some sophisticated developers and admins, that Windows must be rebooted frequently to remain “stable”. Reasons cited are memory leaks, drivers, and unknown errors though I’m sure there are more. I am probably inviting flames myself here but I find that in most cases the people who make these statements fit into one of these categories:
- developers/users who don’t know how to troubleshoot system issues
- system administrators who don’t know how to troubleshoot application issues
- linux system administrators who don’t know who to troubleshoot Windows
The last group probably doesn’t have much motivation to learn either, but that’s another story. I have never experienced these issues with Windows servers, not ever. Most of my experience is after the Windows 2000 releases and currently I am working with Windows 2000, Windows 2003 and Windows 2003 Enterprise versions. The only times we reboot our servers here are when we do system updates that require reboots. This has ranged from weeks to many months at a time and again, the purpose for the reboot was system updates, not performance issues or stability problems. The fact that most Linux upgrades don’t require a reboot is a factor strongly in their favor, but that is really a separate issue. Based on the number of security updates I have gotten from both my Windows versions, and the Debian system I currently use at home, I would say that the number of updates is not a factor that would push you one way or the other. Of course, agnula is not targeted at servers 😉
I have seen memory leaks on Windows servers. I have seen them on Linux servers. They were caused by applications, not by the OS in both cases. In none of the languages that I have written in am I immune to writing leaky code and the issue is identical on both OS platforms. Drivers have a similar profile.
The only people whose opinions I trust on this issue are those who regularly use both operating systems and who don’t have a reason to be biased one way or another. The truth is that there are not many people like that out there. I think I am one of those people. I started with Unix system admin in the early ’90s and have worked with AIX, HPUX, Solaris, FreeBSD, BSDI, Suse Linux 8.2 – 9.1, and Debian verisons of *nix systems. I have administered All versions of Windows since NT 4. All these systems have things that are great and things that stink. It would be really great if instead of spending all this energy on disinformation, suspicious anecdotal evidence, urban legends and downright malicious untruths, we *users* of these systems learned from the best of all of and worked to demand these things from our vendors. That would constitute progress.