Sign in
Microsoft.com
 
Microsoft
 
 
 

locked
Should write caching be turned off on WHS? RRS feed

 > 

  • Question

  • Question
    Sign in to vote
    0
    Sign in to vote

     

    I have just done my 1st build of WHS. I am recycling my old nforce2 shuttle sff. I am booting from an IDE drive connected to the motherboard. I have added a silicon image sata controller and two 750Gb sata drives.

     

    I noticed that the policy for all drives was optimis(z)e for performance with write caching on. Is this recommended for WHS without battery backup (UPS)?

     

    For the moment I have turned off the write caching. I could do this on the IDE drive using device manager. I had to install silicon image's SataRaid tool to turn off the caching on the sata drives. They are not running in raid mode nor do I have the raid bios installed on the sata controller. 

     

    Regards

    Paul

    Saturday, June 14, 2008 10:49 AM

All replies

  • Question
    Sign in to vote
    0
    Sign in to vote

    If you don't have a UPS, and live in an area where you suffer from power glitches, then I would turn it off. You aren't going to see any difference in performance anyway.

     

    Colin

     

    Saturday, June 14, 2008 5:46 PM
  • Question
    Sign in to vote
    0
    Sign in to vote

    I agree. maybe the WHS team should attempt to make the default off or at least mention it in the install guide.

     

    Write caching can be very useful for certain applications. At work, I see great improvements on SQL Server using the write cache on the disk controller. However these servers are all on a UPS and the disk controllers have battery backup.

     

    I'd hate to get a power failure as a vital pointer is being updated.

     

     

    Saturday, June 14, 2008 5:52 PM
  • Question
    Sign in to vote
    0
    Sign in to vote

     

    Just as a final note on this. I turned off write caching on my boot drive (a 2 year old 7200rpm 320GB drive). I did not really notice much difference till I watched it hibernate. From circa 10 seconds this increased to maybe 4 minutes. Very suprising as you would think that the tiny write cache would fill up pretty quick on a sequential write of 1GB. I'm guessing the write cache helps in as a traditional buffer.

     

    Paul

    Tuesday, June 17, 2008 4:57 PM