Windows Server 2019 Launching in October

Megalith

24-bit/48kHz
Joined
Aug 20, 2006
Messages
13,000
Microsoft’s latest version of Windows Server will be reaching general availability next month. Succeeding last year’s version 1709, the latest release reintroduces the desktop GUI and features improved security with Windows Defender Advanced Threat Protection, more efficient containers, Windows subsystem on Linux, and more.

Through various listening channels such as the Insider program, product telemetry analysis, and industry trends, we heard loud and clear that hybrid, security, agility, and TCO are top of mind for our customers. Datacenter modernization is critical to support your business and deliver innovation, especially given the competitive landscape today. Windows Server 2019 is designed and engineered to help modernize your datacenter, delivering on four key areas.
 
I will never understand why all Windows Server versions take FOREVER to start and/or shutdown.

Set up a *nix server doing the exact same thing, and it's as fast as any desktop machine to cold boot.
 
I will never understand why all Windows Server versions take FOREVER to start and/or shutdown.

Set up a *nix server doing the exact same thing, and it's as fast as any desktop machine to cold boot.

I'm running Server 2016 with a low-end quad and it starts as fast as my Ubuntu workstation with the same CPU (both 7600k) and my Windows 10 desktop with an 8700k. And that's with a couple of VMs running in Hyper-V on the server.
 
I'm running Server 2016 with a low-end quad and it starts as fast as my Ubuntu workstation with the same CPU (both 7600k) and my Windows 10 desktop with an 8700k. And that's with a couple of VMs running in Hyper-V on the server.

I didn't set them up myself, but they were all set up by different people. Maybe they're making the same mistakes, but really, it's a PITA.
 
I didn't set them up myself, but they were all set up by different people. Maybe they're making the same mistakes, but really, it's a PITA.

Well, to make any reasonable claim of a difference, you've got to eliminate the variables. I'm a sample of one here and I wouldn't take either of our experiences as hard points of reference, all I can say is that what you reported above isn't always the case. I certainly can't claim that my experience is more representative than yours.

What we'd need is to see thorough testing on common server hardware from say Dell, HP, and Lenovo.
 
  • Like
Reactions: DocNo
like this
I concur with CDR_74, windows server takes forever to boot and shutdown. It depends on services you have, but for the most part its true.
That's one of the reasons I have to schedule updates. So users won't be affected much.
 
For those saying your Windows Server(s) take forever to boot/reboot, can you actually quantify that with some time info? Curious how long you folks are talking about.
 
I will never understand why all Windows Server versions take FOREVER to start and/or shutdown.

Set up a *nix server doing the exact same thing, and it's as fast as any desktop machine to cold boot.
Is your nix box CLI only? or is it gui? (same question for the win server box)

For those saying your Windows Server(s) take forever to boot/reboot, can you actually quantify that with some time info? Curious how long you folks are talking about.
Between two bare metal R720s with the same specs our server 2012R2 box takes about 8 minutes from cold boot going through all post checks, comparably our centos is about 6 minutes I would estimate about 4-5 of that wait is hardware related.
One primary difference though is the windows box is GUI, the centos is a CLI bare, even so, the linux box is slightly faster not by an astronomical margin though.

I have seen many windows boxes with tons of services take substantially longer to startup though.
Even some older 2003 boxes that have issues and/or hardware problems in general I've seen take hours to startup.
The 2008R2 and newer are substantially quicker at their startup times from what I've seen.
 
If you are booting your servers frequently enough to notice their load time, the snarky might point out your doing it wrong :)

The biggest problem I have with Server 2019 is they destroyed the Server Essentials Role. The friendly admin panel was nice, but not critical. What they really killed that is annoying was the built in workstation backup functionality. It is awesome for small businesses/orgs. It has a very smart de-dupe and a bare metal restore and it's so easy to use I have walked people over the phone through swapping a new hard drive into a dead machine, creating the USB recovery key for that computer, booting off the key, and then restoring the computer. It is still a killer feature and one that I will really miss. Indeed, without it it's kind of hard to justify staying on Windows server. I guess after support for 2016 ends I'll have to see if it makes sense to keep fooling around with MS for servers or not. I have one site to get off of SBS 2011 and I guess I'll be moving it to server 2016 for now and then see what's up once support for it ends.

They keep removing features around group policy from Windows 10 Pro up to Workstation that make it worth having active directory around which don't help either. Samba has good enough AD emulation to handle account management and if GPOs are going to be neutered into oblivion unless you pay through the nose for Workstation/software assurance or Windows Enterprise (that requires SA) then screw you Microsoft. I dunno why they are giving up on the SMB market now, when there's even more pressure on traditional desktops from mobile. In enterprise there are already significant pockets of users - especially upper management - that just work off of tablets (iPad Pros 90% of the time). I'm starting to see some noise around the viability of doing the same in some of the SMB sites I support. If Apple ever fixes text manipulation on iOS I could switch to iOS for the majority of what I need to do. Doesn't seem to be a good time for MS to start nickel and dimeing people again but here we go.
 
I will never understand why all Windows Server versions take FOREVER to start and/or shutdown.

Set up a *nix server doing the exact same thing, and it's as fast as any desktop machine to cold boot.
I have no idea what you're talking about. The OS starts just fine unless you're running some junk services.

Are you sure you're not talking about physical hardware POST process?
 
I will never understand why all Windows Server versions take FOREVER to start and/or shutdown.

Set up a *nix server doing the exact same thing, and it's as fast as any desktop machine to cold boot.
I'm running whs2011 on an 1155 Xeon and agree 100 percent
 
Not sure what the shock is, Linux Kernel has always been much more efficient than Windows. It is MONOLITHIC. WIndows is a MicroKernel that has to add in everything as it boots. Also, because of that Windows Kernel design, optimization is secondary.
 
If you are booting your servers frequently enough to notice their load time, the snarky might point out your doing it wrong :)
Do you not ever update your Windows-based servers?
Normally doing so almost always requires a reboot with .Net optimization services running before and afterwards to finish things up.
 
For those saying your Windows Server(s) take forever to boot/reboot, can you actually quantify that with some time info? Curious how long you folks are talking about.
While I haven't used a stopwatch, I could say that some of our servers can taker over 10 minutes to reboot. VM servers can take much longer. Although that makes sense as it has to stop all the VMs.
 
I will never understand why all Windows Server versions take FOREVER to start and/or shutdown.

Set up a *nix server doing the exact same thing, and it's as fast as any desktop machine to cold boot.

You might wanna audit your system and find out what it is hanging on, because that's not normal. For our VMs, the Linux and Windows ones all start up in about the same time. They can be rebooted quite quickly, as you'd hope for VMs. For the few physical systems (mostly hypervisors) it spends WAAAAY longer on the BIOS and iDRAC checks than it does on booting the OS.

So if you have a really slow server, might be worth your while to find out why.
 
Not sure what the shock is, Linux Kernel has always been much more efficient than Windows. It is MONOLITHIC. WIndows is a MicroKernel that has to add in everything as it boots. Also, because of that Windows Kernel design, optimization is secondary.

That is not quite true. Linux adds almost all drivers after the kernel is up and running. They are not statically linked.

There really is no way for a properly configured UNIX box to boot slower than any Windows OS. A clean UNIX box has virtualy no services/daemons running on it. I am pretty sure you c annot get Windows to ever boot that cleanly and that is the major controbutor to the time difference.

Although, like someone mentioned, if boot time is a problem for you, then you are doing something wrong. My backup system (UNIX box) has been up for 5 years, 132 days, 16 hours, as of this second. All my boot drives run RAID 1 allowing me to swap them out in a timely manner.
 
That is not quite true. Linux adds almost all drivers after the kernel is up and running. They are not statically linked.

There really is no way for a properly configured UNIX box to boot slower than any Windows OS. A clean UNIX box has virtualy no services/daemons running on it. I am pretty sure you c annot get Windows to ever boot that cleanly and that is the major controbutor to the time difference.

Although, like someone mentioned, if boot time is a problem for you, then you are doing something wrong. My backup system (UNIX box) has been up for 5 years, 132 days, 16 hours, as of this second. All my boot drives run RAID 1 allowing me to swap them out in a timely manner.

Your UNIX box.

I won't say I am a PRO on Windows Server - not a pro at *nix servers either btw - but as far as I know some updates ask you to reboot, period. I do have an old Xeon (771->775 mod) running CentOS for months without a reboot, being updated twice a day. But the 2012/2016 boxes I've played with recently always ask for reboot after, say, a cumulative update.

In my experience - and it is not a "pro sysadmin" one by any means - Windows Server asks for a reboot every once in a while, and it takes forever. *nix (my experience is mainly CentOS) never asked for it, and when I do reboot it for any reason it goes back up real quick.

That all said, it is very likely there's something wrong with the boxes I stumbled upon, and I also am nowhere near a pro on any of this. But I wasn't the one who setup all the WS boxes, and those were all set up by different people at different times (2008R2/2012/2016). They may be all doing it wrong, sure, n=1 and all, but whenever I see a WS box I know I'll stay there for a while. LOL.
 
That is not quite true. Linux adds almost all drivers after the kernel is up and running. They are not statically linked.

There really is no way for a properly configured UNIX box to boot slower than any Windows OS. A clean UNIX box has virtualy no services/daemons running on it. I am pretty sure you c annot get Windows to ever boot that cleanly and that is the major controbutor to the time difference.

Although, like someone mentioned, if boot time is a problem for you, then you are doing something wrong. My backup system (UNIX box) has been up for 5 years, 132 days, 16 hours, as of this second. All my boot drives run RAID 1 allowing me to swap them out in a timely manner.

We have two ways to have drivers in Linux.

Drivers in the Kernel:
cat /lib/modules/$(uname -r)/modules.builtin

Loadable Kernel Modules LKMs
find /lib/modules/$(uname -r) -type f -name '*.ko*'

Services are a product of the OS and user setup. We can go on and on with that, but for the most part - we are talking "out of the box".

If you want a super fast Linux boot, custom compile the Kernel and leave out what you don't need.
 
We have two ways to have drivers in Linux.

Drivers in the Kernel:
cat /lib/modules/$(uname -r)/modules.builtin

Loadable Kernel Modules LKMs
find /lib/modules/$(uname -r) -type f -name '*.ko*'

Services are a product of the OS and user setup. We can go on and on with that, but for the most part - we are talking "out of the box".

If you want a super fast Linux boot, custom compile the Kernel and leave out what you don't need.

Xeon X3353 + 8GB RAM + CentOS = under 30 sec to boot/shutdown.

2x Xeon E5430 + 12GB RAM + SBS2008 = 15 min to boot/15 min to shutdown.

That's all after BIOS crap, of course. The last one is true server hardware so the BIOS stuff takes a while, but that's expected.

Both running as DNS/DHCP/SMB, and the CentOS one also doubles as a router and cloud storage server (Nextcloud). Service-wise, it would be the one doing the most stuff.

Those are the closest ones I had with hardware that's somewhat comparable. A newer one runs Xeon Silver, Server 2016 and, no joke, took more than an hour to shutdown the HyperV server, and I did shut down all VMs (all two of them) manually beforehand. I am sure that's a glitch, however.

The other one I played with recently was a 2012R2 with a E5 V5. Around 15-20 min like the SBS2008 one. This one was one of the most pointless servers I've ever stumbled upon, as it was massive for... a simple SMB share. LOL.
 
I will never understand why all Windows Server versions take FOREVER to start and/or shutdown.

Set up a *nix server doing the exact same thing, and it's as fast as any desktop machine to cold boot.

I spun up a simple domain with Server 2019 to play around - 2 DCs, 2 file servers, and 1 more for screwing around and all of them except for the PDC start in seconds. As long as there isn't an update pending or something, it's maybe 5 seconds from me hitting reboot to back at the login screen.

Updates in 2016 are completely fucked so I'll give you that. They take _FOREVER_ and it's exceedingly well known that they're absolute shit.

A 15-20 minute boot time is bonkers... something is wrong. I'm around 100~ Windows servers at work and most will happily bounce in a couple minutes on spinners. Are these bare metal or are you bouncing the host itself?
 
All bare metal. The VMs running on the 2016 server that I mentioned taking an hour to reboot/shutdown go off fairly quick. WTF, indeed.
 
I would nuke a server if it took longer than 30 seconds to boot. All my UNIX boxes (FreeBSD, Solaris, Linux) all come up in less than 15 seconds. All of them are CLI based. In my opinion, GUI's get in the way of proper server administration in a UNIX environment,

The only time I have ever had a box take a minute or more, is when any one of the RAID mirrors had to be re synched with the master. Even then, it is still less than 1.5 minutes.

Shutdown is almost instantaneous.

We have two ways to have drivers in Linux.

Drivers in the Kernel:
cat /lib/modules/$(uname -r)/modules.builtin

Loadable Kernel Modules LKMs
find /lib/modules/$(uname -r) -type f -name '*.ko*'

Services are a product of the OS and user setup. We can go on and on with that, but for the most part - we are talking "out of the box".

If you want a super fast Linux boot, custom compile the Kernel and leave out what you don't need.

Well, of course. Who runs a stock box? I certainly would not. Whenever I see a UNIX box (really does not matter what flavor) I assume the admin is smart enough to have tailored it to the operations it was hosting.

Although there have been rare occasions when I did see a UNIX system configured badly for what it was supposed to be doing (i.e. listeners on ports not needed, services running not needed....).
 
Do you not ever update your Windows-based servers?

At least once a month!

Normally doing so almost always requires a reboot with .Net optimization services running before and afterwards to finish things up.

I guess we don't have the same .Net dependencies or just have faster hardware. A few machines where I can't afford to move to SSD I have used PrimoCache and an SSD to speed them up and it can be quite dramatic if your OS isn't on an SSD already. Native Windows disk caching is almost worse than useless :/
 
Also I find whenever any edition of Windows starts to take forever on boot or shutdown there is something up with DNS. Just about every single time. Even when I don't think it's DNS, if come back and do some more digging - yup, DNS wasn't working like I thought it was.
 
Back
Top