Padre, or anyone, ever heard of this?

[G+_John Sullivan]

Padre, or anyone, ever heard of this? A guy in a Yahoo group that I subscribe to had a network of four computers that ran fine until he had to replace the network switch.

After that, transfer speeds between computers were horribly slow. After much research he fixed it by going into the Advanced tab of his network adapters and setting both LargeSendOffload IPv4 and LargeSendOffload IPv6 to Disabled.

Anybody run across that before? thanks

[G+_Travis Hershberger]

While a little odd, that's not an abnormal problem to have. Hyper-V is known to do the same sort of thing (make network speeds very slow.)

 

Back in the late 90s (showing my age here), we had all sorts of problems with auto-negotiated network speeds. Our 100mb networks were only connected at 10mb, that took a bit to figure out as it was a bug with SGI network adapters and the 3com switches being used.

[G+_Shooter_FPV (Shooter_FP]

John Sullivan I did, about maybe 6-7 years ago. I think it was called TCPIP Offload, or something like that. Same situation though. Didn’t happen often, and I never figured out what the trigger was. And I believe that it would happen to maybe 2-3 systems on the same network, not all of them like your situation.

 

I know this might not help, but I’ve seen it before...

[G+_Paul Hutchinson]

I'd never heard of this, this article gives a good description of what it does and why it can go wrong causing a slowdown.

peerwisdom.org - Large Send Offload and Network Performance | Peer Wisdom

[G+_Brent Vrieze]

Paul Hutchinson, very nice article. So sum up shutting off the setting puts the onus back on the CPU so the MTU size is known. One way to fix this would be to just set your switch for "jumbo frames" and let the LSO (see article posted above). I now where I work we used to set every switch to jumbo frames but we have now suspended that setting.