File server backup using 3PAR VSS to StoreOnce slow in comparison to MSL


DP Version: A.09.08: OMNIDBCHECK, internal build 113

CM: Windows 2012 R2 , on DL 380G6, 1 CPU 24GB RAM

HPE 3PAR VSS Agent: A.09.08

File Server: Windows 2012 R2, VM with 4CPU & 32GB RAM

StoreOnce: 3540, 4TB SAS7.2K RPM disk

When I perform backup of my Fileserver on DISK using HPE 3PAR VSS agent it gives me an average backup spped of 230GB/hr while the same backup on MSL gives me 450GB/hr.

Is there any way I can increase backup speed on DISK? As per my knowledge it may be becuase of dehydration and slow disk but wanted to take second opinion.

Thanks & Regards,


  • Hello Syed,

    Looks like this is a Performance-related query and so this may need to be investigated further - starting with checking your current configuration.

    When I perform backup of my Fileserver on DISK using HPE 3PAR VSS agent it gives me an average backup spped of 230GB/hr while the same backup on MSL gives me 450GB/hr.

    As you know, we cannot compare apples and orange, and so same thing, I dont think we can compare this two backup - one to disks and the other one to MSL.

    Better you concentrate on the one you have an issue - which I believe is your backup to disks.

    Is this a new configuration ? Has this been working before - have you had the chance to get the backup speed you need ?

    How are things configured with your filesystem backup ?

    Looks like a lot of things may need to be checked and as such, usually performance related issue is handled by DP consultant as he/she may be able to advise what is the best configuration, specific to your environment and your enterprise backup requirements.

    Hope this helps



    HPE DP Support


  • Hello,

    Sherman is more than right.

    to start with focus on how you tranfer your data to SO.

    Which devices did you create and via where are they reached.

    If you f.i. write to a catalist over the 1GB interfaces of your SO you should even thank your SO to be so quick :-)

    My golden rule here: follow the path of your data........ and you will see


  • Hello Sherman & Luc,

    Thanks for your reply.

    I am using catalyst stores for my backup on Disk with medium as CoFc.  I am using this backup for almost a year but on MSL and the speed is same from day 1 till today. We have purchased a new StoreOnce (3540) with catalyst license 2 months back and since then I am taking my FS backup first on DISK & then on MSL.

    I am using 8GB SAN with HPE 3PAR VSS Agent for this backup.

    I am using MSVSSW backup jobs to perform this backup. As far as backup speed is concerned I am ok with it if no other options is there but looking for any chance of improvement to decrease my backup window.

    The performance is same on both DISK (230GB/hr) & if taken directly on MSL (450GB/hr) from day one.

    Thanks & Regards,


    Thanks & Regards,


  • CoFC,  which gateway?

    and if this gateway is the same host as the one who's writing to the MSL,

    what are the connection stream limits settings on your specific HW/stores/gateway

    and what's the activity which occurs simultameously on your SO , cause the appliance has its limits......

    Once more: follow your data path and at each step check the limits  (SAN/appliance/DP logical definition)

    And is the blocksize definition within your logical device definition of DP similar?


  • CoFC GW is our CM which is the same GW for backup on MSL also.

    There is no connection stream limits settings on HW/Stores/GW all are default.

    The activity occuring on SO is other parallel backup with this backup but I have tried performing this backup alone with no other jobs on SO have same performance.

    Block Size is set to 512 for both SO & MSL.

    Not sure how to check data path and at each step check the limits  (SAN/appliance/DP logical definition)

    Thanks & Regards,


  • How many CoFC devices does your host see ?  == number of devices per login in SO GUI

    This is comparable with "concurrency"  of a tape device.


    If your number of data-streams differs, your data-amount will differ as well

    Check this out as well please


    Luc Minnaert

  • under device manager of my CM which is also the GW for this backup I can see 128 HP StoreOnce CoFC SCSI Processor Device.

    Number of devices per Login in SO GUI shows 64 for each FC port (FC Port 1 = 64 devices, FC Port 2 =64 devices)

    Thanks & Regards,


  • And now we touched a soar spot !!

    To my humble knowledge, 64 is much to much.

    64 per fabric == 128 devices; compare this with a server having to feed 128 LTO drives .....  !!!

     by the way what kind of server is this? if your responding a Windows than i will start laughing . for this kind of setup you need a dedicated superdome running HP-UX

    and even then, your SO is doing in-line dedupe. Do you think this machine can handle 128 parallell streams simultaneously.

    ==> you created a CPU & memory performance issueon host and/or StoreOnce appliance

    and that explains the difference

  • What should be the ideal Number of devices per login? Having 128 devices does it makes that 128 LTO drives are writing at the same time? As per my understanding this is the maximum number of drives that I can used which I can never reached. 

    I have tried running this backup alone which means only 1 drives is used for writing then too having same performance.

    I remember earlier number of devices per login was 32 with no differences in the performance.

    This server is DL380 G6 single processor with 22GB RAM having windows 2012 R2.

    Performance of GW server: CPU= Average 50%; Memory Average =40%

    SO in last 24 hrs shows: CPU Node1: Max utilization of 25%; Memory NOde1: Max utilization of 45%

    This results is during my weekend when full backup going on with max parallel streams .

    Thanks & Regards,



  • Verified Answer

    Hello Syed,

    you would have to dive into the release notes of your SO FW release to be able to make an estimated guess.....

    1 thing is sure, Your server is undersized to deliver top performances (my laptop has 8 CPU's and 16 GB RAM (but still is too slow since it too runs Windows))

    But why does it work better with tapes? Well, tapes were and are always under-estimated (with propper tuning I always can make tapes faster :- )  ) 

    A thing we didn't mention yet:  compare the blocksize; Probably your tape device is writing in a bigger / more suitable blocksize than your SO-logical device....................


  • Thnaks Luc for all your effort and help.

    I feel I need to close this thread here as I feel now it's the limitation of my CM rather than SO.

    FYI blocksize is same for both tape & SO set as 512.

    Thanks & Regards,


  • My Pleasure


    Luc Minnaert Technical Consultant

    Technology Services

    Enterprise Group 32 2 620 37 75  Office Personal Info Erased Mobile

    Personal Info Erased

    Hewlett-Packard Belgium BVBA/SPRL

    Hermeslaan 1A

    B-1831 Diegem

    RPM/RPR Brussels


Reply Children
No Data