Using Windows 7 Ultimate and Windows Server 2008 R2 Two machines, connected via 1GB 1000base-T LAN, with a 1Gb switch between... and yes, I've tried swapping cables and ports. Server has two RAID5's, a 6 platter and an 8 platter. Workstation has two RAID5's, a 3 SSD and a 4 platter. Both sides using Intel CPU's and a RealTek 8100C chipset Gb LAN, same LAN drivers. Using Windows Explorer to COPY a 20GB file: Server Disk1 to Server Disk 2: >250MB/s (Server disk I/O isn't an issue). Server Disk2 to Server Disk 1: >250MB/s (Server disk I/O isn't an issue). Workstation Disk1 to Disk2: >150MB/s (Workstation disk I/O isn't an issue). Workstation Disk2 to Disk1: >150MB/s (Workstation disk I/O isn't an issue). Server Disk1 to Workstation Disk2: 91MB/s (near 100% LAN utilization). All looks good so far: Workstation Disk2 to Server Disk2: 25MB/s declining to 17MB/s; occasionally as low as 9MB/s. Some of the changes listed above appeared to work for a few moments. I've seen as high as 90MB/s for a few seconds, but it declines rapidly under 30MB/s... so no fix that has lasted. It appears there is something in Windows 7 that prevents fast transmits; or something in Windows Server 2008 that blocks fast receives. I have set the server-side receive training to disabled, normal, highlyrestricted, and experimental and the Win7 side to off, normal, and highlyrestricted in various combinations. Again, nothing concrete to report as an improvement. What would the MS Gurus suggest I do with NetMonitor as far as capturing something worth their time to diagnose?

Hi, You may use the Network Monitor to capture the packages and to see whether can get anything of value. Based on my understanding, this method may be able to find out the root cause. Also, try the following. Disable Auto-Tuning http://support.microsoft.com/kb/934430 How to use the throttling mechanism to control network performance in Windows Vista http://support.microsoft.com/kb/948066 Hope this helps.Please remember to click Mark as Answer on the post that helps you, and to click Unmark as Answer if a marked post does not actually answer your question. This can be beneficial to other community members reading the thread.

I've tried disabling auto tuning on the clients with no appreciable effect. The throttling feature however appears to have had an effect - changing the setting from default 10 to 70 has slowed the transfer from 20MB/s down to about an initial 6MB/s that has slowed even further, and is now at 3.2MB/s (3% network utilization according to Resource Monitor at both ends of the wire - WS2008R2 and W7). Question: What does the 10 (0x0a and 70 (0x46) represent: Mbps on the LAN, or 10% and 70% reserved, or? I'll try the "off" setting of all FF's and add to this. In any case, having a backup server that can't be utilized due to client "entertainment media" issues is a little painful... there should be a straightforward and published method to set the network to "normal". UPDATE: I've tried settings of 0, 10, 70, and FFFFFFFF. 10 might have been best, but only marginally better than 70. Range of time to copy was under 7 minutes to over 8 minutes. Very small difference. But I also tried this using Windows Explorer and a Dataset of one folder containing 12.5GB of miscellaneous folders and files - and was surprised. - Dataset is at \\WIN7\Dx\ = = Client physical RAID5 drive D:\Dataset On WIN7 Client: Share drive D with Everyone as Dx On WSvr Server: Share drive D with Everyone as Dx On WIN7 Client: Map \\WSvr\Dx to Z: On WSvr Server: Map \\WIN7\Dx to Z: On WIN7 Client, using Windows Explorer: -- Copy Dataset from Local Drive D: to Network Drive Z: -- Took about 8 minutes, with 20% network utilization. On WSvr (2008 R2 Server), using Windows Explorer: -- Copy Dataset from Network Drive Z: to Local Drive D: -- Took about 3 minutes, and 70% network utilization. Why does running the copy from the server allow better network utilization and net throughput?

I've tried disabling auto tuning on the clients with no appreciable effect. The throttling feature however appears to have had an effect - changing the setting from default 10 to 70 has slowed the transfer from 20MB/s down to about an initial 6MB/s that has slowed even further, and is now at 3.2MB/s (3% network utilization according to Resource Monitor at both ends of the wire - WS2008R2 and W7). Question: What does the 10 (0x0a and 70 (0x46) represent: Mbps on the LAN, or 10% and 70% reserved, or? I'll try the "off" setting of all FF's and add to this. In any case, having a backup server that can't be utilized due to client "entertainment media" issues is a little painful... there should be a straightforward and published method to set the network to "normal". UPDATE: I've tried settings of 0, 10, 70, and FFFFFFFF. 10 might have been best, but only marginally better than 70. Range of time to copy was under 7 minutes to over 8 minutes. Very small difference. But I also tried this using Windows Explorer and a Dataset of one folder containing 12.5GB of miscellaneous folders and files - and was surprised. - Dataset is at \\WIN7\Dx\ = = Client physical RAID5 drive D:\Dataset On WIN7 Client: Share drive D with Everyone as Dx On WSvr Server: Share drive D with Everyone as Dx On WIN7 Client: Map \\WSvr\Dx to Z: On WSvr Server: Map \\WIN7\Dx to Z: On WIN7 Client, using Windows Explorer: -- Copy Dataset from Local Drive D: to Network Drive Z: -- Took about 8 minutes, with 20% network utilization. On WSvr (2008 R2 Server), using Windows Explorer: -- Copy Dataset from Network Drive Z: to Local Drive D: -- Took about 3 minutes, and 70% network utilization. Why does running the copy from the server allow better network utilization and net throughput? UPDATE2: I am NOT using HOMEGROUP; I have a dedicated workgroup network the client is a member of, and the server is a DC. I'll eventually try joining the client to the domain and see if that changes anything, but doubt it will so haven't taken the time yet.

Problem remains and has not gone away, so please do not mark something as an answer just because you haven't or can't solve this issue.

Interesting read and might help: http://blogs.technet.com/b/askperf/archive/2007/05/08/slow-large-file-copy-issues.aspx Not so much the eseutil but the new switch for xcopy ( /J Copies using unbuffered I/O. Recommended for very large files. ) Hope that helps.MCSE/MCSA/MCP/Security+,A+

Turkey - thanks. While that provides annother alternative, the slowness isn't in the file i/o, rather, its in the network send (or perhaps network receive side on the server). Since the process of backing up files from clients to servers is something everyone should want to do, it seems odd to me that this issue doesn't have a technote specific to maximizing the network utilization during these types of operations. Plus, allowing the clients to do the work offloads the server of that task. I've resorted to using robocopy on the server pulling files from the clients. This seems to provide best utilization. Though makes recovery of a failed client more tedious.

Have you solved the problem? I seem to be getting a similar situation

No resolution yet - I haven't researched this, it might a default server GPO regarding non-domain client resource limitations, one that prevents clients from using more than 10% network bandwidth.

No resolution yet - I haven't researched this, it might a default server GPO regarding non-domain client resource limitations, one that prevents clients from using more than 10% network bandwidth. I've had some weird problems in my home network - large files transfer well, but small files and any kind of video playback from my WHS 2011 server were terrible - lots of pauses and basically unwatchable. The problem turned out to be the WHQL Realtek drivers. I upgraded to the latest drivers from the Realtek website (dated 23/5/11) and the problem went away.