none
DPM & ESEUTIL causing long running jobs RRS feed

  • Question

  • I am using a single DPM server (with asecondary) to backup 4 Exchange servers (CCR pairs) each with 5 storage groups/mailbox around 150GB each.

    I am doing an express full backup nightly on the 20 mailboxes and to be fair everything is going fine however  ESEUTIL appears to be a very slow process when running more than 2 protection processes.  This causes some impact on job completion and results in some very long running jobs.

    Question is :

    Is this normal behaviour for ESEUTIL and are there logs relating to the ESEUTIL process stored anywhere on the DPM server?

    Would I be safer increasing  the number of DPM servers say to 1 DPM Server per 2 Exchange clusters ?

    Thursday, December 30, 2010 5:21 PM

Answers

  • Hello,

    It does scale but it's all relative to the resources in use. There are a lot of variables in place when talking about performance.

    Example: If your Exchange has 2 SG's that are 20 gigs and DPM has just a handful of protection groups that replicate at night only and the bandwidth is decent, then offloading eseutil check on DPM shouldn't be too much of a problem.

    If your Exchange has 5 SG's each around 150 gigs and you have many PG's that replicate all throughout the day and each server is set to use compression on the wire which adds to the CPU hit and the page file is off for DPM and antivirus is aggressively scanning the directories and the bandwidth is the heavily saturated etc.... <I could go on and on> then offloading eseutil would be slow.

    You may have some or none of the things above in place. The most common "gotchas" are:

    DPM page file is incorrect.
    Antivirus scanning the DPM directories.
    Synch jobs not staggered.

    I have never seen DPM perform a eseutil check on SG's faster than what a local Exchange eseutil can do. This is especially true if you have all the SG in the same PG and are trying to have them all synch at the same time.

    Questions:
    If you run the eseutil on the local exchange server does it complete much faster than if the DPM server were to perform the eseutil check? If so, then by how much?
    I would expect it to be somewhat faster as DPM is not just performing the eseutil check but also has two move over the changed blocks across the wire but the question would be is the difference huge.
    Example: Exchange finishes a eseutil check in 45 min and it takes DPM 5 hours for the same SG. If so, then I'd start looking at the resources in play vs. the DPM setup.

    If Exchange finishes in 45 min and DPM takes 1hr. 15min, then I wouldn't be concerned.   During this test though the playing field needs to be considered.  If exchange is a quad proc, with 32 gigs of memory naturally it'd be much faster on a single eseutil check on one SG than if DPM was installed with the minimum specs running a synch on all the SGs at the same time while trying to perform an eseutil check.  

    Articles that may help you.

    DPM Managing Performance: http://technet.microsoft.com/en-us/library/ff399471.aspx
    How DPM Operations Affect Performance: http://technet.microsoft.com/en-us/library/ff399315.aspx
    DPM and Memory: http://technet.microsoft.com/en-us/library/ff399244.aspx
    Performance Counters: http://technet.microsoft.com/en-us/library/ff399366.aspx
    Improving Performance: http://technet.microsoft.com/en-us/library/ff399571.aspx
    Managing DPM Performance on a WAN: http://technet.microsoft.com/en-us/library/ff634226.aspx
    How Protection Groups Changes Affect Jobs: http://technet.microsoft.com/en-us/library/ff399316.aspx
    DPM and Antivirus: http://technet.microsoft.com/en-us/library/ff399439.aspx

     

    Thanks
    Shane

     

     

    Friday, January 21, 2011 2:42 PM
    Moderator

All replies

  • Actually after doing a bit more analysis I retract my statement above that "everything is going fine" and have come to the conclusion that offloading ESEUTIL activity for more than one backup process at a time stinks.  DPM for Exchange on a single server does NOT scale.
    Thursday, December 30, 2010 8:37 PM
  • Hello,

    It does scale but it's all relative to the resources in use. There are a lot of variables in place when talking about performance.

    Example: If your Exchange has 2 SG's that are 20 gigs and DPM has just a handful of protection groups that replicate at night only and the bandwidth is decent, then offloading eseutil check on DPM shouldn't be too much of a problem.

    If your Exchange has 5 SG's each around 150 gigs and you have many PG's that replicate all throughout the day and each server is set to use compression on the wire which adds to the CPU hit and the page file is off for DPM and antivirus is aggressively scanning the directories and the bandwidth is the heavily saturated etc.... <I could go on and on> then offloading eseutil would be slow.

    You may have some or none of the things above in place. The most common "gotchas" are:

    DPM page file is incorrect.
    Antivirus scanning the DPM directories.
    Synch jobs not staggered.

    I have never seen DPM perform a eseutil check on SG's faster than what a local Exchange eseutil can do. This is especially true if you have all the SG in the same PG and are trying to have them all synch at the same time.

    Questions:
    If you run the eseutil on the local exchange server does it complete much faster than if the DPM server were to perform the eseutil check? If so, then by how much?
    I would expect it to be somewhat faster as DPM is not just performing the eseutil check but also has two move over the changed blocks across the wire but the question would be is the difference huge.
    Example: Exchange finishes a eseutil check in 45 min and it takes DPM 5 hours for the same SG. If so, then I'd start looking at the resources in play vs. the DPM setup.

    If Exchange finishes in 45 min and DPM takes 1hr. 15min, then I wouldn't be concerned.   During this test though the playing field needs to be considered.  If exchange is a quad proc, with 32 gigs of memory naturally it'd be much faster on a single eseutil check on one SG than if DPM was installed with the minimum specs running a synch on all the SGs at the same time while trying to perform an eseutil check.  

    Articles that may help you.

    DPM Managing Performance: http://technet.microsoft.com/en-us/library/ff399471.aspx
    How DPM Operations Affect Performance: http://technet.microsoft.com/en-us/library/ff399315.aspx
    DPM and Memory: http://technet.microsoft.com/en-us/library/ff399244.aspx
    Performance Counters: http://technet.microsoft.com/en-us/library/ff399366.aspx
    Improving Performance: http://technet.microsoft.com/en-us/library/ff399571.aspx
    Managing DPM Performance on a WAN: http://technet.microsoft.com/en-us/library/ff634226.aspx
    How Protection Groups Changes Affect Jobs: http://technet.microsoft.com/en-us/library/ff399316.aspx
    DPM and Antivirus: http://technet.microsoft.com/en-us/library/ff399439.aspx

     

    Thanks
    Shane

     

     

    Friday, January 21, 2011 2:42 PM
    Moderator