I very much implore anyone who is part of the decision process of server software acquisitions to avoid using Windows Server if you have the choice. It is painful to work with over many years and has been outclassed in many ways by free and open alternatives that are easy to adopt.
I say this as someone who has been responsible for server clusters for over a decade and has had the unfortunate fate of maintaining critical business infrastructure on Windows Servers.
The open source options are still a bit like a kit car, while Windows is an awkward but finished product. And I say that as a FreeBSD and NetBSD developer and a career primarily building high scale Linux infra. If you can use some type of PaaS/FaaS perhaps the concern approaches zero since you are again using some more finished product.
The one area I think Windows can gain some market is HyperV, with VMWare becoming expensive. I recently landed a bunch of fixes to FreeBSD network drivers so they SR-IOV mode works. It's an odd hypervisor, but it seems to work in my limited testing.
Microsoft got that right - they determined what people actually wanted, then created a product that did that. Nobody is doing as well in Linux, not even Red Hat. Should we start today?
TBH I've not used it professionally but Samba actually has done a lot of work building up a usable AD DC platform that works on linux. It punts a lot of work over to LDAP and co generally but you can get substantial parity to ADDS on linux nowadays.
OIDC generated a credential. That credential isn’t going to instantly expire for zero-trust clients. You’ll generally have anything from 8 hours to several days before needing to re-auth against your identity provider.
Just like with Windows logins against AD. Once you’ve had an initial login from your laptop, you can continue to log into your laptop for days without having a connection to AD.
To counter this, while usually a non-Windows server can be better, there are plenty of use cases where Windows is the best solution and you should use it in those circumstances. Evaluate every deployment individually and pick the best tool for the job.
Maintaining a Windows server removes most of the headaches of long term Windows server use, and it's also the area most people don't keep up with causing the issues to start.
In many ways it hasn't been outclassed. Every time I dip a toe into the Linux waters, I recoil in horror at the things that lurk beneath the surface.
You're all suffering from Stockholm syndrome.
Where to begin?
There is no Linux, except the kernel. There's dozens of distros, and none are even vaguely compatible in their user space in the same way Windows is. Docker containers are effectively mandatory if you want to retain your sanity.
Configuration files are ad-hoc, unspecified, text formats (huuurk) with arbitrary and capricious escaping and quoting rules. It's rare for these to be formally specified anywhere. Good luck doing anything to Linux configuration programatically!
A Linux operating system in general is "write only, once" -- which also explains the popularity of docker. Trying to manage a fleet of heterogeneous servers with Ansible, Puppet, or whatever is an exercise in frustration. They're barely more than script-execution engines, applying opaque changes that are difficult to audit later. You're always best off discarding servers entirely and starting from scratch.
Linux barely even has a concept of a "service", which invariably require a full stop-start to pick up configuration changes. When -- finally -- systemd was introduced to catch it up to NT4, there was much wailing and gnashing of teeth. It's still inconsistent, so publishing server software for Linux is again, you guessed it, best done with containers.
Don't even get me started on how Linux stumbled into UTF-8 in an era where Windows, MacOS, and Solaris had already been doing Unicode for a decade.
Can I put spaces in file names yet, or is that still considered an eldrich incantation to greybearded admins?
Unicode in user names?
Central auth without hard-coding IP addresses for "the" (singular) Kerberos server, like it's an MIT lab in the 1980s?
Here's a trivial operation on Windows that would make any Linux admin collapse into a twitching heap and foam at the mouth: Make every server and workstation trust the same private root certificate authority, for every HTTPS client on the system, not just one particular web browser or app runtime.
Or... ensure the existence of a consistent firewall rule to every Linux computer, e.g.: for Tenable scanners.
Or... add a local user to every Linux computer with some non-trivial rights.
PS: I've done all of the above and was horrified about the time, complexity, and cost.
anybody that decides "windows server " should be part of their infra stack in the past 20 years clearly has ulterior motives (ie, kickbacks); and has likely never touched/managed a windows server
I wish they didn't discontinue the free Hyper-V server, last release was in 2019. Yes Proxmox or xcp-ng (Xen) are better choices nowadays but if you have some Windows VMs or legacy Windows software then I think Hyper-V was/is pretty cool (in fact it doesn't matter because you can run Linux, BSD any OS inside the hypervisor it doesn't matter).
I have a homelab PC running it, the actual only big downside is the lack of ZFS support. Otherwise it's still a good solution (in a 'not great but works perfectly' sense)
I'm going to try to stay on topic: what are the practical use-cases for the folks here who are running Windows Server, to continue to do so? Momentum? Legal?
I work in the cybersecurity space and I see Windows Server everywhere. More common with smaller shops (think ~50 or less people). HN tends to skew very tech forward and isn't aware of the breadth chaos that's out there. A lot of the times this is because smaller companies don't do their own IT. Think of your dad/aunt trying to use a computer. Those same people sometimes end up running business that need servers. They don't manage the IT but go through a MSP for it. MSPs universally tend to be Windows shops that want to do things as cheaply and quickly/conveniently as possible. That means they push Windows and their clients don't know better to push back. It's also probably convenient for the end user. They don't need to think about server/computer. Their mental model can become "this is my computer, and that's the computer for the printer to work", because Windows server looks similar enough no one bats an eye. If you're not a technical person Windows server is just Windows and so you can work with it.
We had a client (manufacturing) that still used server 2008 running some kind of access app. It hadn't been updated or even touched in years and was the cornerstone of the business for them. Until I saw it myself I thought thinks like that were only stories. I've seen them a lot at some smaller union funds/offices. They'll just have one server that's used for sharing/printing/etc. I'll bet if you go to any medium sized city in America (say Cincinnati) and find a local wealth advisors office (or other similar local mom and pop shop) they've probably got one as well. Also doctors. For some reason every smaller doctors practice I've every worked with (about 9 now) love using Windows Server. There's just so much stuff in the med space that just works with Windows only.
Part of it's probably a very healthy dose of inertia. It works so there's no reason to spend time (and more often money) to change it.
For people/places like this I think Windows Server is a good option. It just works out of the box, integrates well with stuff most people know, and can be managed with a GUI so it's not super scary for the non technical folks.
This. Around 2020-ish, I went for a interview at small EU IoT startup that had about ~10 employees, and the founder and CEO told me how happy he was they went all-in on the MS stack since they just pay for Office 365 & Azure subscriptions and everything cloud and IT related is basically taken care of via his Macbook, no need for on-prem infra or an army of expensive sys-admins to manage a Linux stack. He said he manages everything in the company from one account on his macbook and for him that's a game changer so he can focus on the product, business and sales instead of maintaining and micromanaging a tech stack.
And this is besides all the banks, insurance and finance corps here that have 20+ years of legacy stacks all running on MS or on their way transitioning to MS after decades of IBM(so much for EU tech sovereignty).
I have no experience in running or configuring any of these massive MS stacks, but it seems whatever MS does it pleases the people who are paying for the product, and in those business that's the main selling point. HN PoV seems to be biased towards the user and UX rather than the beancounters.
Everything seems right until the last sentence; banks, insurance, finance, a small start-up (that may or may not be tech focused, I'd guess not), they aren't getting Windows because it's a downgraded UX for their purposes. And they aren't all (or even any in some cases) bean counters.
They are getting _something_ that they're happy about, and they're not all blind to alternatives.
The reason why anyone runs any OS: because it runs the software that we need.
Some of that software might be dependency of some other software which is a dependency on some other software deployed everywhere.
Almost any organization is a complex series of integrations -- almost nothing is it's own silo that doesn't communicate with another system or application.
Cooperative development on Windows. That's how we've used it. We're on 2012 R2 still. We run a very, very narrow set of tools and don't uses browsers on the machine.
We updated a 2nd machine to 2016, but with SSDs the performance was as bad as the 2012 R2 machine with spinners. OLD spinners.
I spent 2 days looking into how to purchase a more modern version... honestly, I was shocked how confusing it was. That was a couple of years ago.
Typically Active Directory, and organizational inertia (existing software is meant to run on Windows Server, knowledge/contractors/etc. are familiar with WinServer) as you mentioned. Sometimes, certain applications are only certified for compliance on Windows Server. Small businesses seem to run Windows Server because it comes with a UI, usually because IT is an off-the-side of the desk thing for someone.
In big enterprises it's fairly pervasive for all the imaginable reasons - the ubiquity of Office / Exchange to run your business, and the build out through Active Directory and outwards to properly capture your IT.
FYI / fwiw and on a more specific note, we use them as RDP to run a large number of software and utilities only available in Windows form. For example, the Peoplesoft ERP can and does run on Linux and aix, but the proprietary GUI IDE is windows only. Our dbas use Toad. For whatever reason our RPA Blueprism is running on windows as well.
A. We need to run Windows programs. Our 3D cad PDM is windows only. Our ERP is windows only. Our camera system and access controls, also windows. We need to work with other businesses, and guess what? They use windows too. It’s pure fantasy that you can run many businesses without Windows. Maybe self-contained ones that sell software or services directly to customers. I’m in automotive; it’s all windows.
B. Our MSP like almost all of them is a windows shop. It’s what they know how to use and they can train employees on it. We could hire internal IT, but it would be at least double what we’re paying our pretty good MSP and we don’t need to deal with staffing. We aren’t an IT company and don’t want to be.
Personally, I won’t use windows ever again. But, most of our people have no choice. Put it down to institutional momentum if you must.
Honestly because it’s cheaper than Enterprise Linux: buy CD keys of Windows Server perpetual licenses for like $100 rather than $900/year subscription per CPU for something like RHEL.
I find Windows to be more stable, too. RHEL kept randomly dropping fiber optic adapters every few weeks, which would require network manager to be restarted, and nobody knew why. Stock standard Dell PowerEdge with certified RHEL, supposed to be rock solid but not as stable as expected
When we used RHEL we used to hit the support like 2 or maybe 3 times per year, mostly issues with things breaking during upgrades.
Since switching virtualisation hosts to Windows Server (we still run Linux VMs), performance has been measurably the same, updates have been fully automatic with zero intervention, and management has been simple — we haven’t needed any support for the last 5 years.
I was pretty surprised at WS stability as a virtualisation host because everything online said HyperV was crap, but it’s actually pretty good. Our workloads are HPC: we hold all the cores at 90% nearly 24/7.
Mostly because Microsoft is a huge company and thus it's perceived as more trustworthy than the alternatives. Imagine the average CEO looking at the Windows Server webpage vs the Ubuntu Server webpage. The Ubuntu page looks comical by comparison.
C-suite and execs know it, the software the customer is running on it is proven stable for their use case (inertia as opposed to momentum), and maybe the people deciding on things get intimidated by figuring out which other OS to use. Do you use Red Hat? Canonical? Do you give a child's eternal soul over to Larry Ellison as a part of the contract for Oracle Linux? If you just stick with Windows Server, for better or worse, you don't have to answer these questions.
There isn’t a good reason to be running Windows Server today, besides quirky legacy compatibility requirements.
If it requires user auth integration, Windows is really the best choice still for high compliance environments where user auth stems from AD. But I’d argue organisations deploying servers that still need user auth are doing something drastically wrong today.
Windows is a well paved road for legal and compliance, but so is RHEL.
GNU/Linux, FreeBSD, OpenBSD, NetBSD are way way much better choices. There are no reasons today to go with a licensed, proprietary and closed source product. Make the switch and don't ever look back.
I’ve always used Windows Server for development machines (since NT 3.5) but this sounds like that may be a bit problematic now? A shame as I am using Windows 11 at work right now and .. ok I don’t hate it. But there is plenty I don’t like.
I say this as someone who has been responsible for server clusters for over a decade and has had the unfortunate fate of maintaining critical business infrastructure on Windows Servers.
The one area I think Windows can gain some market is HyperV, with VMWare becoming expensive. I recently landed a bunch of fixes to FreeBSD network drivers so they SR-IOV mode works. It's an odd hypervisor, but it seems to work in my limited testing.
I'm no Windows-lover, but usually ADDS is the thing enterprises are after with Windows Server infrastructure.
Leveraging an AD Domain makes life a lot simpler when your workforce isn't mostly developers and engineers.
If you're not in a big cloud environment though, I do agree. ADDS is hard to beat.
Just like with Windows logins against AD. Once you’ve had an initial login from your laptop, you can continue to log into your laptop for days without having a connection to AD.
We're all reliant on many third parties to different extents to keep our services online.
I'll have fun with modern IAM and cloud network security tooling.
Maintaining a Windows server removes most of the headaches of long term Windows server use, and it's also the area most people don't keep up with causing the issues to start.
Skill issue.
This sentence implies you were the open-source IT guy who thought "oh it's just Windows I can figure it out".
Being a good Windows admin takes skills and experience, it's no different than e.g. Red Hat.
You're all suffering from Stockholm syndrome.
Where to begin?
There is no Linux, except the kernel. There's dozens of distros, and none are even vaguely compatible in their user space in the same way Windows is. Docker containers are effectively mandatory if you want to retain your sanity.
Configuration files are ad-hoc, unspecified, text formats (huuurk) with arbitrary and capricious escaping and quoting rules. It's rare for these to be formally specified anywhere. Good luck doing anything to Linux configuration programatically!
A Linux operating system in general is "write only, once" -- which also explains the popularity of docker. Trying to manage a fleet of heterogeneous servers with Ansible, Puppet, or whatever is an exercise in frustration. They're barely more than script-execution engines, applying opaque changes that are difficult to audit later. You're always best off discarding servers entirely and starting from scratch.
Linux barely even has a concept of a "service", which invariably require a full stop-start to pick up configuration changes. When -- finally -- systemd was introduced to catch it up to NT4, there was much wailing and gnashing of teeth. It's still inconsistent, so publishing server software for Linux is again, you guessed it, best done with containers.
Don't even get me started on how Linux stumbled into UTF-8 in an era where Windows, MacOS, and Solaris had already been doing Unicode for a decade.
Can I put spaces in file names yet, or is that still considered an eldrich incantation to greybearded admins?
Unicode in user names?
Central auth without hard-coding IP addresses for "the" (singular) Kerberos server, like it's an MIT lab in the 1980s?
Here's a trivial operation on Windows that would make any Linux admin collapse into a twitching heap and foam at the mouth: Make every server and workstation trust the same private root certificate authority, for every HTTPS client on the system, not just one particular web browser or app runtime.
Or... ensure the existence of a consistent firewall rule to every Linux computer, e.g.: for Tenable scanners.
Or... add a local user to every Linux computer with some non-trivial rights.
PS: I've done all of the above and was horrified about the time, complexity, and cost.
We had a client (manufacturing) that still used server 2008 running some kind of access app. It hadn't been updated or even touched in years and was the cornerstone of the business for them. Until I saw it myself I thought thinks like that were only stories. I've seen them a lot at some smaller union funds/offices. They'll just have one server that's used for sharing/printing/etc. I'll bet if you go to any medium sized city in America (say Cincinnati) and find a local wealth advisors office (or other similar local mom and pop shop) they've probably got one as well. Also doctors. For some reason every smaller doctors practice I've every worked with (about 9 now) love using Windows Server. There's just so much stuff in the med space that just works with Windows only.
Part of it's probably a very healthy dose of inertia. It works so there's no reason to spend time (and more often money) to change it.
For people/places like this I think Windows Server is a good option. It just works out of the box, integrates well with stuff most people know, and can be managed with a GUI so it's not super scary for the non technical folks.
And this is besides all the banks, insurance and finance corps here that have 20+ years of legacy stacks all running on MS or on their way transitioning to MS after decades of IBM(so much for EU tech sovereignty).
I have no experience in running or configuring any of these massive MS stacks, but it seems whatever MS does it pleases the people who are paying for the product, and in those business that's the main selling point. HN PoV seems to be biased towards the user and UX rather than the beancounters.
They are getting _something_ that they're happy about, and they're not all blind to alternatives.
Some of that software might be dependency of some other software which is a dependency on some other software deployed everywhere.
Almost any organization is a complex series of integrations -- almost nothing is it's own silo that doesn't communicate with another system or application.
We updated a 2nd machine to 2016, but with SSDs the performance was as bad as the 2012 R2 machine with spinners. OLD spinners.
I spent 2 days looking into how to purchase a more modern version... honestly, I was shocked how confusing it was. That was a couple of years ago.
FYI / fwiw and on a more specific note, we use them as RDP to run a large number of software and utilities only available in Windows form. For example, the Peoplesoft ERP can and does run on Linux and aix, but the proprietary GUI IDE is windows only. Our dbas use Toad. For whatever reason our RPA Blueprism is running on windows as well.
B. Our MSP like almost all of them is a windows shop. It’s what they know how to use and they can train employees on it. We could hire internal IT, but it would be at least double what we’re paying our pretty good MSP and we don’t need to deal with staffing. We aren’t an IT company and don’t want to be.
Personally, I won’t use windows ever again. But, most of our people have no choice. Put it down to institutional momentum if you must.
That way, in server I user power shell and in desktop bash! how should be!
Would love to have switched off of Windows but we run to many legacy applications.
I find Windows to be more stable, too. RHEL kept randomly dropping fiber optic adapters every few weeks, which would require network manager to be restarted, and nobody knew why. Stock standard Dell PowerEdge with certified RHEL, supposed to be rock solid but not as stable as expected
Since switching virtualisation hosts to Windows Server (we still run Linux VMs), performance has been measurably the same, updates have been fully automatic with zero intervention, and management has been simple — we haven’t needed any support for the last 5 years.
I was pretty surprised at WS stability as a virtualisation host because everything online said HyperV was crap, but it’s actually pretty good. Our workloads are HPC: we hold all the cores at 90% nearly 24/7.
If it requires user auth integration, Windows is really the best choice still for high compliance environments where user auth stems from AD. But I’d argue organisations deploying servers that still need user auth are doing something drastically wrong today.
Windows is a well paved road for legal and compliance, but so is RHEL.
[1]: https://www.windowscentral.com/microsoft/windows-11/microsof...
Cool. Any plans to not have the number 4 issue by design?