When you look at traffic stats for a physical interface input is obviosuly from the downstream device and output is what is going to the downstream device. (I am basing this from a core switch perspective).
With a VLAN I take it that output traffic shows the total going to ALL ports within the VLAN and this may include trunk ports from which a VLAN has nto been pruned and will of course include the etherchannel between 2 core devices. Similarly for input traffic, that will be coming from many sources, WAN, user stcks and so forth. I am also of the view that the figures would include traffic between devcies within the VLAN, say two servers talking to each other.
This would explain why when I graph the data for my production server VLAN I see peak values of 120%, against what I take to be a notional 1 Gbps - given its 6509 - with 4 blades of 1 Gbps ports.
or am I way off beam here?
Cheers