Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

We are deploying physical windows servers (460c/Gen8/128GB RAM/300GB HDD) if I implement the page file, following the best practice of 1.5 times the ram, its going to take up the complete disk almost, so with such a large amount of RAM will it be necessary to have a page file that follows the normal windows way.

share|improve this question
    
I've a server with 128 RAM (windows 2012) that hosts 16 VMs, and the page file is automatically managed by the system, and it only takes 20GB. What are you going to use that server for? –  EliadTech Jun 25 at 13:25
    
What best practices do you quote that go for 1.5 times ram? Yes, I remember that being a best practice a long time ago; but not the case anymore. –  TomTom Jun 25 at 14:05
    
I don't believe that the "rule of thumb" you're following is applicable to Windows Server 2008 and forward. Have a read of the article at the link. Personally, with W2K8 and forward I set the pagefile to system managed. - blogs.technet.com/b/markrussinovich/archive/2008/11/17/… –  joeqwerty Jun 25 at 14:20

3 Answers 3

up vote 5 down vote accepted

There are a few things to consider here.

  1. If you don't have a pagefile the size of RAM + 257 MB, you cannot get a full crash dump. If you set it too low, you can hit OOM conditions.

  2. You can also leave it at "system managed" and measure what the system is going to do with it.

  3. The 1.5x RAM recommendation is old and no longer makes sense. I would throw that out of the window and forget it ever existed.

So now you have to ask yourself:

Do you care about a full crash dump? No, probably not. Minidumps are fine in almost all cases and very few people take the time to actually analyze or understand crash dumps.

Can you forecast memory utilization in your system? If you can, you can estimate how large to set the pagefile to with some extra space built in as a cushion. I'm guessing that if you could do this, you wouldn't have asked this question.

Is the answer to the previous two questions "no"? If so, just go with System Managed and keep an eye on it, because you'll get yourself in trouble otherwise.

share|improve this answer
2  
1 x RAM plus 257 MB –  Mathias R. Jessen Jun 25 at 14:40
    
Whoops. Good call –  MDMarra Jun 25 at 14:41

the short answer is: no, such a big swap file is not needed.
anyway this is an highly debated question.
reading this page you will find the best practice you are referring to, a prominent disclaimer stating that this best practice is not coming from microsoft and reading the comments you will discover that there are people referring to swap files as 90's heritage...

as suggested by microsoft documentation check the requirements of any application you are going to deploy, try to estimate memory usage based upon expected user/work load, decide if you want/need full memory dump and act accordingly.
you may even discover that a 16gb swap file is ok in your setup with your hardware configuration.

share|improve this answer

Just to clear up the various "Rules of Thumbs" here they are in summary:

Note: anywhere I say 1x RAM, I actually mean 1x RAM + 260MB. See KB2860880 for details

For Server 2000 and 2003:

  • For small amounts of RAM, about 2-4GB, the page file should be 1.5x RAM.
  • For large amount of RAM, the page file should be the same size as RAM.

For Server 2008 and 2008 R2:

  • For "normal" servers let Windows manage the page file. It will almost always be 1x RAM.
  • For Hyper-V only host servers, set the page file to 8GB (unless you want a full memory dump on crash, then set it to 1x RAM).

For Server 2012 and 2012 R2:

  • Let Windows manage the page file in all situations.
    • In "normal" servers it will tend to be 1x RAM (unless you have a ton of RAM, then it's usually less than 1x, in any case Windows will figure out what it needs and you don't have to fuss with it).
    • In Hyper-V only host servers it will normally be a few GB (mine are ~3GB right now). Hyper-V only needs about 500MB of swap to operate normally (slightly more with management and monitoring software), the rest is space for a kernel dump should the server blue screen.
  • if your server has blue screened in the last 28 days, Windows will up the page file to just over 1x RAM so that on the next crash it can capture a full memory dump (essential for debugging the most pesky blue screens)
share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.