5

To be clear: I'm not asking for opinions, rather I'm looking for data/facts/insights that will help me get closer to forming my own opinion.

Context

We're about to upgrade our servers (as they're old), and our current supplier is proposing to put Windows Server 2016 on the new server.

I've been told that it might be a better idea to keep running on 2008r2, since it's a proven and stable version, as opposed to 2016 which is new. Also, our company isn't that big (50 computers), so we don't need cloud/hybrid stuff that's in 2016.

Question

I've read about Microsoft's Premium Assurance, which allows using 2008r2 beyond its 2020 end-of-life date.

There's also upgrading to 2012r2, which has an end-of-life date of 2023.

Upgrading to 2012r2 seems like the best option, given Microsoft even still allows buying 2012r2 licenses. I'm also wondering about how and if the licensing system is different between 2008r2 and 2012r2.

Can someone provide some insight that will help me make/get closer to a decision?

6
  • 3
    What is your reason for not doing it?
    – eckes
    May 15, 2017 at 14:33
  • 2
    There's surprisingly little real difference between 2012R2 and 2016 for >70% of users - so certain you should try to go to 2012R due to 2008R2's EoSL issue but if you're putting in the effort to do that then you may as well go straight to 2016 to get the extra years of service life.
    – Chopper3
    May 15, 2017 at 14:37
  • 2
    The fact that this is a small company is a very big reason to upgrade to 2016.
    – EEAA
    May 15, 2017 at 14:46
  • 3
    There isn't enough information about your environment to provide a useful answer. Also, the licensing is not really a factor. You always purchase the current version of the license, and the Microsoft license allows you to install and run the previous version. So you can purchase 2016 licenses, install and run 2012 R2 if you need it, and upgrade to 2016 later.
    – Greg Askew
    May 15, 2017 at 14:53
  • 2
    There are a lot of threads regarding choosing between 2012R2 and 2016, but this one is my favorite one, it may be helpful. reddit.com/r/sysadmin/comments/5fbwg3/…
    – Stuka
    May 17, 2017 at 9:40

3 Answers 3

2

It's hard to answer your question. You won't find public insights that tells you you should upgrade to the "not latest" Windows Server. I guess because Microsoft's influence in the IT press is very high and they always push their latest system, and so it's hard to find someone telling you the other way around.

As insight, i can share my 20 years experience on Microsoft OSs. The latest is never been a good option. The only exception was upgrading Vista to 7, for who went to Vista (I had all my clients still to XP ofc). Someone says Microsoft likes to use their users as beta testers, but many people may argue on this. Someone say you should wait between 1 to 2 years before installing latest OSs.

For instance, with latest updates from last patch tuesday we experienced on a few 2016 Std this error: "The stub received bad data". This just happens the first time you open up an application in the desktop environment. If you click the program icon again, it just normally opens. Now, such a weird thing never happened on a "at least two years old" microsoft os. This is just an example.

Waiting a few years is always been the best option with MS Servers as for my experience. The only one thing that should lead you to the latest software is if there are unbackported features you desperately need.

Also, you talk about a small env, and as thus i imagine you don't have a test env. So implementing latest softwares will be even more risky.

Also, consider what ppl wrote in the comments. Some of them won't aggree with me. Licensing: purchase 2016 Std. Run 2012 R2. When it comes to the next server upgrade, that will probably be before 2012R2 EOL (2023), you will find a good chance to upgrade to 2016. Again, you'll find yourself asking if you should upgrade to 2020R5, or to 2016. You'd better choose 2016, or, at least, that's what i would do.

5

Going 2008R2 -> 2012R2 in mid-2017 is pretty much pointless, I'd wait for 2016 successor (there are rumors MSFT will release it soon to tap numerous holes in their WS2016 GA) and put IT into production.

1

I pretty much second Marco's experience. In February for instance, I tried to setup a Windows Server Failover Cluster with Storage Spaces Direct for an SQL Failover Cluster instance. The way to achieve this was very rocky and last week, the whole thing deadlocked and crashed in on me. Finding proper, detailed and relevant documentation on Windows Server 2016 features is slowly getting better now, but still a major pain. And I ran into a lot of not so funny stuff feeling a bit like a beta tester. In this case, I had no choice due to the required feature set, but my experience holds true that the latest version is not necessarily the most mature one.

So if you just need a plain server for "normal" duty, I don't see a reason not to go to 2016. We deployed several, it will give you the platform and increase the interval before the next upgrade. But if you have something more taxing in mind, I'd check whether 2012 R2 delivers all features you need.

3
  • 3
    We have another story to tell: S2D production collapsed after one node was put down for maintenance. May 22, 2017 at 15:16
  • 3
    From what I've seen, such issues happen all the time. Having played around s2d, we ended up with starwind vsan. It does HA storage and doesnt suffer from the degraded perormance if one node dies. Just for your info, it is a platform agnostic, so you can safely consider it as a decent alternative to s2d. May 23, 2017 at 16:28
  • 3
    I'd give Microsoft another year or two to catch up. Original Storage Spaces in WS2012 were pretty much unusable (especially parity modes) up to 2012R2 where they got major improvement. S2D is V1.0 technically. May 23, 2017 at 20:14

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .