none
Express full vs Consistency check VM via Hyper-V host-RA

    Question

  • What is the difference between an express full backup (recovery point job) and a full consistency check when backing up online VM's from a host-based RA on Hyper-V 2008 R2.

    According to documentation, when using a VM-based RA, it will use the NTFS change journal to identify changed blocks and that optimization will allow jobs to run faster and with a lower impact on the protected resource.  But, does a host-based RA still use the NTFS change journal to identify changed blocked in hosted VHD files?

    I ask because while it seems that a consistency check does seem to take roughly twice as long as a recovery point (via express full backup) using host-based RA (indicating that they are in fact doing different things), it still seems that a recovery point of a VM from host-based RA takes considerably longer than it should if it actually used the NTFS change journal "shortcut" to identify blocks.

    Can anyone describe in detail how each works specifically with a host-based RA protected Hyper-V 2008 R2 VM's?

    EDIT: Using DPM 2010 with RU 2.

    Thanks in advance!


    Monday, June 06, 2011 3:14 PM

Answers

  • If the above suggested answer does not help in resolving the thread please re-open it.
    --------------------------------------------------------------------------------
    Regards, Michael V [MSFT] - This posting is provided "AS IS" with no warranties, and confers no rights.
    Thursday, January 19, 2012 6:30 PM

All replies

  •  But, does a host-based RA still use the NTFS change journal to identify changed blocked in hosted VHD files?
    When I posed this question, I meant does the host-based RA use the NTFS change journal on the host to identify changed blocks that make up the raw VHD files.  I assume that beyond the initial Hyper-V VSS Writer informing the writers within VM of the pending backup so that it makes the VHD consistent and quiet in time for the host VSS snapshot, that the host-based RA itself does not "reach into" the VHD at all and treats it as one large bag of bits.  Thus, I would not expect the host-based RA to utilize the change journal of the VM, only the host itself.
    Tuesday, June 07, 2011 1:47 PM
  • HI,

    As you have eluded to, DPM does not use the file system filter driver to track block level changes to the .VHD file(s) when doing host level protected, DPM uses a targeted CC against the VM's .vhd(s) and only brings over changed blocks.   This is the reason the host level backup times are proportional to the size of the VHD's associated with the protected guest.

     


    Regards, Mike J. [MSFT] This posting is provided "AS IS" with no warranties, and confers no rights.
    Tuesday, June 07, 2011 3:42 PM
  • Mike,

    Thanks for the reply.

    So then how are the express full backup and consistency check different?  It seems like they must both do the same thing, which is to scan the entire VHD at both ends, comparing hashes and/or CC's along the way in order to detect changes, and sending those blocks across the wire.  But, are they different at all in the case of backing up VM's and VHD's from a host?

    Tuesday, June 07, 2011 4:07 PM
  • HI,

    Under the covers there is no difference how changed blocks get detected and brought over, the only difference is after the changes are applied, an express full will make a recovery point, a CC will not.   Hyper-V protection is the ONLY workload that uses CC logic for express full backups, that is because block level changes are not exposed to our DPM filter driver if the VM is located on CSV cluster disk so we needed to use CC logic.  We will make this better in the next version of DPM.


    Regards, Mike J. [MSFT] This posting is provided "AS IS" with no warranties, and confers no rights.
    Tuesday, June 07, 2011 4:13 PM
  • HI,

    Under the covers there is no difference how changed blocks get detected and brought over, the only difference is after the changes are applied, an express full will make a recovery point, a CC will not.   Hyper-V protection is the ONLY workload that uses CC logic for express full backups, that is because block level changes are not exposed to our DPM filter driver if the VM is located on CSV cluster disk so we needed to use CC logic.  We will make this better in the next version of DPM.


    Regards, Mike J. [MSFT] This posting is provided "AS IS" with no warranties, and confers no rights.

    Hmm.  So if my Hyper-V host is a standalone with DAS and not clustered, then DPM is tracking block changes via its filter driver and shouldn't have to scan the full VHD for recovery points?  That would explain why express full backups seem to be shorter than consistency checks I suppose.  Still it seems like if it were tracking block level changes it would go faster than it does and have less performance impact on the DAS storage.

    Tuesday, June 07, 2011 4:58 PM
  • If the above suggested answer does not help in resolving the thread please re-open it.
    --------------------------------------------------------------------------------
    Regards, Michael V [MSFT] - This posting is provided "AS IS" with no warranties, and confers no rights.
    Thursday, January 19, 2012 6:30 PM