none
DPM back-end IO write size of 64KB... and page file usage... RRS feed

  • Question

  • Hi,

    1a) I have been informed that the disk (DPM storage LUN) IO payload size (disk data transfer block size) (i.e. whilst reading/writing the VSS snapshots) is 64KB for both reads and writes.

    1b) Or have I been mis-informed and what my source actually meant... is that the 'disk volume blocking factor' of the storage LUN for DPM is 64 KB, i.e. that the many GPT partitions which comprise used DPM storage space have a blocking factor of  64KB ?

    1c) Or, are both statements above true?

    2) If (1a) above is true, then is there a way to increase the IO payload (disk data transfer) size, and thus move more backup data per IO ?

    3) I've also been told that a DPM server instance might also typically require a large amount of page file space.  Can someone explain why this might be the case?  Could this be overcome by providing more RAM?  Or is there something inherent in the design of DPM as an application that simply requires 'N' x RAM for page file space?

    4) The requirement of a large amount of page file space would seem to indicate to me that the page file will experience lots of activity, and thus that the page file for a DPM server instance (physical or virtual) should itself be placed upon the fastest possible storage.  True?

    Thanks.

    Thursday, December 11, 2014 3:28 PM

All replies

  • bump.
    Tuesday, December 16, 2014 4:37 PM
  • Hi,

    Yes, the DPM agent is optimized for 64K IO and is not configurable.  Windows will automatically align volumes on 1MB IO boundaries which is evenly divisible by 64K.

    Page file requirements are for Windows to maintain VSS snapshots as per the following TechNet article.

    Scalability Factors for Shadow Copies

    You can get a DPM page file calculator powershell script from the following post.


    Please remember to click “Mark as Answer” on the post that helps you, and to click “Unmark as Answer” if a marked post does not actually answer your question. This can be beneficial to other community members reading the thread. Regards, Mike J. [MSFT] This posting is provided "AS IS" with no warranties, and confers no rights.

    Tuesday, January 6, 2015 8:23 PM
    Moderator
  • Hi,

    Yes, the DPM agent is optimized for 64K IO and is not configurable.  Windows will automatically align volumes on 1MB IO boundaries which is evenly divisible by 64K.

    Page file requirements are for Windows to maintain VSS snapshots as per the following TechNet article.

    Scalability Factors for Shadow Copies

    You can get a DPM page file calculator powershell script from the following post.


    Please remember to click “Mark as Answer” on the post that helps you, and to click “Unmark as Answer” if a marked post does not actually answer your question. This can be beneficial to other community members reading the thread. Regards, Mike J. [MSFT] This posting is provided "AS IS" with no warranties, and confers no rights.

    Is this still true for DPM 1807 with ReFS filesystem?  I'm looking for confirmation on the optimal way to setup the storage pools on a dedicated server with internal storage, non-virtualized.  Should I use 64KB stripe element size on the RAID controller (mine defaults to 256KB) and format ReFS with 64KB Bytes Per Cluster?  I have 12 disks to work with, so I'm thinking of doing a pair of RAID 5(s) -  5 disks for each array with 2 global hot spares.  Would those settings be acceptable?
    Monday, January 7, 2019 5:52 PM