I am having a similar problem, is this problem solved at your site? If it is
please inform me about the solution.
Thanks
Luit
"(E-Mail Removed)" wrote:
> Good Morning All!
>
> I, along with several colleagues have been banging our head against a
> brick wall for the last few days, trying to get to the bottom of an
> odd networking (i think) problem.
>
> We have a number of Windows 2003 Servers (currently running SP1).
> These servers were built a while ago, but they are just having their
> final configurations applied prior to going live.
>
> During some data copies, it became apprent that there was something
> strange going on with file transfer times. A 450Mb file took around
> 17 minutes to copy, we retried that copy again, and this time it took
> around 25 seconds (about what it should). This happens on a number of
> servers on site.
>
> At this point it's probably worth going into a bit of detail about the
> enviornment. All the servers are IBM X Series servers, they all have
> dual gigabit network cards teamed. The switch infrastructure is made
> up of Cisco 3750 x 4 in a single stack. There are no firewalls,
> routers, bridges or anything else on this particular network. We get
> the file copy problem when the source and destination servers are
> within the same switch, and split between switches.
>
> The servers have 4gb of RAM, and dual 36Gb disks configured as RAID 1.
>
> Since that initial copy we've conducted hundreds of file copies across
> the network, between different servers, and the results are totally
> inconsistent. One minute the copy will take 25 seconds, the next it
> will take 18 minutes. We've also had a few fail with the "Network
> name no longer available message" (not many though). At the time the
> files are being copied, there are no drops from the ping tests (and
> everything is <1ms as expected). Even when we had the network name no
> longer available messages - the pings didnt drop.
>
> We've made quite a few changes / tweaks to test things, so i'll list
> those now:
>
> Stopping Trend AV on both Source and Destination server
> Checking that all ports in the network team were running at 1000/FULL
> Testing with ports set at 100/FULL
> Configured TCP Chksum offloading (both on and off)
> Updated System BIOS
> Updated NIC Firmware
> Updated NIC Drivers
> Updated NIC Teaming Software
> Disabled SMB signing via Local Group Policy
> Disabled LDAP signing via Local Group Policy
>
> Installed (at Microsoft's request)
>
> KB925903
> KB936192
> KB931304
>
> Modified the following registry keys (again at Microsoft's request)
>
> HKEY_LOCAL_MACHINE\System\CurrentControlSet\Servic es\LanmanWorkStation
> \Parameters
>
> Value name: ReadAheadGranularity
> Data type: DWORD Value
> Radix: Hexadecimal
> Value data: 16
>
>
> HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Servic es\MrxSmb
> \Parameters
>
> Value name: InfoCacheLevel
> Data type: DWORD Value
> Radix: Hexadecimal
> Value data: 10
>
>
> HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Servic es\lanmanserver
> \parameters
>
> Value Name: SizReqBuf
> Data Type: DWORD Value
> Radix: Decimal
> Value: 65535
>
> In summary, we are still no further forward in identifying the root
> cause of this issue. Microsoft are continuing to go through 1.5Gb of
> network traces captured on both successful and failed file copies, but
> thought I would bounce it around these groups to see if anyone has any
> bright ideas that might help???
>
>
> Many thanks in advance!
>
>
> Bruce
>
>
|
Similar Threads
Linear Mode
