The most stable version of Windows EVER, in my experience, was NT 3.51 SP5 - and by MY standards I don't class that as "properly stable for a server" but closer than any other Windows version since.
NT 4 SP 6 was fairly close.
2000 never got as stable as NT 4 but was sorta in the same ballpark eventually.
XP ... stable ..... don't make me LAUGH. NEVER was stable enough to run a server on for anything serious, and IMO was iffy for WORKSTATION usage from a reliability standpoint.
You do know that all Windows kernels (except for the Win95/98/ME branch) derive from the original NT kernels, right? Stability has only gotten better. And using XP as a "server" is still a perfectly valid solution for many small businesses. The 2003 kernel is rock-solid if it's not used as a workstation.
The "server" versions seem to have improved some vs 2k, but haven't had AS MUCH experience with those - and I STILL have never seen anything in Windows on a server manage more than months of uptime, with the exception of ONE NT 3.51 SP5 server I saw manage almost 18 months of uptime before it flaked for the first time, and never more than a year after that.
Then I think you are exaggerating. Where I work currently, we have still have an old ERP system that runs on Win 2000. It just runs. The only time we have to reboot it is if we have a full data-center outage.
LINUX I have owned more than a few servers that went 15+ YEARS of "no downtime except for hardware failure, power outage longer than the UPS battery could last, or had to shut them down to move them" and more than a few that had 4+ YEARS of continuous uptime between moves/power outages of excessive length.
One place I worked had a LINUX file server that had not been rebooted in almost 10 YEARS while I was working there (NCS had both a UPS setup AND a genset that interfaces with the UPS for longer outages when I was working there).
Yep, and there are plenty of examples of Windows servers doing the same. Single-function servers that are never tweaked or experimented on will run until the power is shut off. That's the same with linux or windows.
On PRODUCTION controlling computers, Windows is a rarity IME - and ALWAYS had issues with low reliability where it WAS used as anything other than a "monitor what the dedicated hardware doing the ACTUAL WORK is doing" interface or as a "terminal to talk to the mainframe/server that's doing the ACTUAL work".
Well, now you are talking about a different class of compute. I agree that mainframes are a different breed and you won't find much windows there... the hardware architecture doesn't support it.
Had one place replaced a QNX box with a Windows-based "upgrade system" from the same company - and after 2 months of nothing but problems and frequent downtime that was costing quite a bit of productivity, along with *3* visits from a FACTORY TECH to try to make it work right, the company put the QNX box BACK in place because it WORKED - despite being 10 year older hardware it worked FASTER AND MORE RELIABLY than the Windows-based machine that was supposed to be an UPGRADE.
Yes, I like QNX, we used it as the base OS for our video content server when I worked for a company back in the late 90's. We sold our systems to hotels and to large telecom giants in Asia to deliver MPEG2 video to set top boxes for video on demand applications. The reason we used QNX was for it's real-time OS abilities, as a blip in MPEG2 video was worth the cost of not having the servers bog down due overloading. We used Solaris for the controllers, a big reason was because that's what the programmers were comfortable with. We had NT controllers in the lab that ran with the exact same reliability. We could charge a lot more for the Solaris boxes, so it became a business decision to do so.
And yes, far too many "Enterprises" know that Windows is going to crash on them occasionally, and have work-arounds in place for when it does so - or they move their servers to a LINUX (or in some cases back in the day a UNIX) solution they can count on to just keep going and going and going and going....
I'm not here to bash linux. It's rock solid too. Any enterprise worth their mettle will have contingency plans, no matter the OS.
There's a reason the MAJORITY of web servers run on LINUX and have done so for a couple decades or so (BSD was the leader before that, in it's various flavors, after taking over from VMS).
The only reason Windows is even in that competition is too many IT shops have lots of Windows experience but NO LINUX experience at all, and many other IT shops don't believe you CAN "mix and match" successfully.
No, the reason is because it's hard to get paid-for support for linux as compared to windows. It's gotten better over the last 10 years, but to suggest that no enterprise ever needs help with a linux based system due to bugs or instability is just fallacy.
Web servers are run on linux based systems due to market momentum. The most popular web server software and tools started there... just like the majority of desktops run windows due to market momentum.
Hint - do you think Google runs on a Microsoft solution?
Answer - no, they run their servers on their own customized LINUX version, and their standard in-house desktop OS is a modification of Ubuntu LINUX.
Well that's a specious argument... why would they? Google doesn't want to pay their competitor for licenses or support.
This is not speculation, this is FACT from having worked with the stuff for decades as a tech and software tester.
I tend to disagree with your argument.
Have you ever been in one of the Azure server centers?
Yes, I've toured the San Antonio location to see whether their vGPU solutions could work properly for our CAD/CAM related needs. We toured SpaceX and Lockheed Martin to see how they are using vGPU/VDI solutions to better equip their engineers. I've also toured a Google data center in North Carolina and an AWS facilty in Virgina. I do this for a living.
I have - Quincy is quite close to where I live and I know a couple of the techs that work there.
Azure as a platform is reliable - but only because it's designed for massive redundancy and fast fail-over when an individual server flakes out - and that's straight from techs that WORK with the infrastructure behind Azure.
In the 3 years that we've been using Azure based servers, we've never had to reboot a server due to "instability". That's the point. Windows as a server is rock-solid reliable.
Any argument to the contrary is just old-school projecting.
MOSIX based platforms can easily match the reliability of Azure.
No doubt. That was never the question for discussion.