Sign in
 
HomeWindows Server 2012Windows Server 2008 R2Windows Server 2003LibraryForums
Windows Server TechCenter >
> 
> 
Slow file copy INBOUND 2008R2x64

Answered Slow file copy INBOUND 2008R2x64

  • Monday, August 13, 2012 4:49 PM
     
     
    Sign In to Vote
    0

    I have read a bunch of other threads on this issue, but no solution seems to work...

    The problem is copying files TO a 2008r2x64 server - Two new servers - both fresh installs and clean - all updates done

    If I copy a file FROM the 2008 r2 servers to an older 2003r2, it is as expected, but inbound I get about 160Kbs – Brutal

    A 20Mb file I am testing with takes a couple of minutes inbound - outbound it is instant – literally

    Copy between the 2 new 2008r2 is also as slow, but I am assuming that it is the INBOUND on the Rx side that is the issue regardless

    I have ruled out the infrastructure (wire/switches) and am using known good connections (ones that function perfectly with 2003r2)

    The servers are both Dell R510 - same rev. - NICs are embedded Broadcom dual 5716C 1Gb - firmware/drivers current (as of 3 days ago at least)

    I have tried disabling the offload features - Chimney, RSS, NetDMA and autotuning - I have set the adapters to full/half, 1Gb/100Mb etc (and all combinations)

    Both servers have the Hyper-V role - one uses a dedicated NIC, the other is shared with the VM Switch (which I did because of this issue - to speed up the speed between the host and guest - the Virtual switch is BLAZING fast)

    I looked at the articles regarding Hyper-V specifically, but they don't fit the situation....

    None of the adapters show any collisions (zero) and all pass internal diags (for what that is worth)

    Infrastructure issues show show up regarless of direction - and would have affected my 2003r2 servers when I moved them to he wire/ports that were originally used by the 2008r2 - (I even transfered the patch cables to test, although my FLUKE shows all wire is OK)

    I am down to NIC firmware or driver - and OS –

    (Broadcom only supports OEMs - Dell - well - they have some 'person' that runs the diag util, says the card is OK - and when they pass me up the chain, they keep dropping me off hold - then BACK again to the same thing - and although they have tel# and e-mail more than once... no callback - and of course I haven't got to someone that will give me there number/ext yet - MOST annoying - sorry for the vent)

    Again, to be clear - absolutely NO special config here - the wiring is safely ruled out - so is the switch - and the offload/chimney, RSS, autotune etc etc have all been tried in many many different combinations...

    Any help or ideas would be wonderful - Cheers....

    ssampson


     

All Replies

  • Tuesday, August 14, 2012 7:31 AM
    Moderator
     
     
    Sign In to Vote
    0

    Hi Scott,

    Thank you for the post.

    I suggest you first update SMB driver files with latest version via install KB2563210 and KB2612966.
    http://support.microsoft.com/kb/2563210
    http://support.microsoft.com/kb/2612966

    Moreover, try to stop VMs and disable VM NIC on both servers to test the copy file speed.

    If there are more inquiries on this issue, please feel free to let us know.

    Regards

    Rick Tan

    TechNet Community Support

     
  • Tuesday, August 14, 2012 1:37 PM
     
     
    Sign In to Vote
    0

    Cheers -

    I will look at those 2 articles now...

    FYI - I have tried turned off all VM services and the associated NICs....and tested using the one server that does not share the NIC with the virtual switch 

    Another note - I found this running a network trace - on an INBOUND file copy...(just the part I found interesting)

    Connection doesn't appear to be bandwidth limited (B/W test #2)
    Analyzed 380 data packets and 14117 ack packets
    Added 14116 outstanding data measurements
    Flight count: 378
    minFlightTime = 102, maxFlightTime = 107532
    minFlightSize = 45, maxFlightSize = 167
    meanFlightSize = 64
    Connection doesn't appear to be bandwidth limited (B/W test #2)
    minRtt = 82, maxRtt = 107532
    retransCount = 0, pureAckCount = 4130

    I am going to try an outbound capture and compare the RTT and Flight - I am guessing they will be lower....

    I MAY have ruled ot the SWITCH prematurely - it is an older switch and perhaps is creating issues for the much newer NICs - Last firmware for the Switch was 2009.... (I did that update months ago - first weekend onsite) Not sure when the actual switch was installed - new IT staff and I didn't support them for 7 years while out of the contry - another consultant)

    Anyway - Thanks for the response - I will check those SMB articles and run the other capture and post then (likely after 5pm EST)

    Cheers

    Scott

    ssampson

     
  • Tuesday, August 14, 2012 3:05 PM
     
     
    Sign In to Vote
    0

    Hey Rick...

    I just finished reading the KBs you sent...I have requested them, but before I install, just a quick query...

    The symptoms listed are very different (actaully, although issues with SMB 1.0 and 2.0 has been something that I wondered about....

    I will check the paged pool usage when I do another capture later....

    So - are you giving me those hotfixes in just in general as a test?? (i.e. the article content is not relevant) or is it because you expect one of those symptoms....

    Just curious - because I am now doing this remotely (not by choice - hehe) and reboots are stressful via RDP  ;-)

    Cheers

    ssampson

     
  • Wednesday, August 15, 2012 5:59 AM
    Moderator
     
     
    Sign In to Vote
    0

    Hi Scott,

    Installed these hotfixes could narrow down the issue if is caused by SMB drivers. Some case logs show these two hotfixes works in their scenarios.

    Regards

    Rick Tan

    TechNet Community Support

     
  • Thursday, August 16, 2012 1:21 PM
     
     
    Sign In to Vote
    0

    Thanks Rick -

    Just like to know the reasoning sometimes !!!

    I will get on that over the weekend.... A major issue has cropped up at another client  (GDC services not starting).... so I must deal with that before I start back...

    Sorry - no new info for now...

    Thanks Again

    Scott

    ssampson

     
  • Monday, August 20, 2012 1:14 PM
     
     Answered
    Sign In to Vote
    0

     

    Well.... not so dramatic a solution as one would like - It WAS the switch after all - in combination with the specific NICs
     
    An older linksys/cicso branded SML2048 (‘final’ - EOL - firmware release) and Broadcom 5716s...
     
    No collisions or anomalies recorded by the switch or the NICs –

    But if I move the connections to other ports, everything began to function at an acceptable speed (no physical wire changes – just switch port)
     
    However – an Intel 1Gb NIC in the port works fine – so initially when I had been looking for a ‘known good’ port to rule out the switch, I had used a port that liked the Intel, but it seems not the Broadcom
     
    I checked only 20 ports and of that found 6 that were OK with the Broadcom NICs – all 20 work fine with the Intel –

    And the whole time I never saw a frame error, collision etc on anything – NIC or switch..
     
    Of course while the problem has been isolated, the conditions are still unacceptable – so the switch MUST be replaced
     
    I am sorry I didn’t get to try those hotfixes Rick (for feedback purposes) –

    This was an accidental discovery - we moved a port to fit a bulky snag hood on a patch cable..
     
    Thanks for your help…
     
    Scott

    ssampson