I've put together a test network in order to test that when an uplink from a 3560 to a 6500 switch saturated by traffic my QOS configuration reserves some bandwidth for management protocols, e.g telnet/ssh.
This is because I have a potential oversubscription model that may be used in production and I need to still be able to manage the switches when the trunk is overwhelmed by customer traffic.
My problem is that even though I have a couple of linux servers which are gig connected to the 3560 pumping out up to 800Mb/s of UDP packets each the trunk only passes up to about 975Mb/s, and this is before implementing any form of QOS.
I presume the 3560 is buffering/queuing/dropping packets, though I don't see any evidence of that from 'show interface' on either the server ports or the trunk port, so I'm still able to ping and manage the switch.
Can anyone shed any light on this? Do I therefore not need to worry about QOS as the trunk will never get fully saturated?