You can support a form of rate-limiting through the implementation of QoS. Things to remember are to disable flow control on all the switch ports (including the gig ports) and to enable qos global:
conf t
int range f0/1 - 24
no flowcontrol receive
exit
mls qos
end
Then create the class map and the policy map:
conf t
class-map match-all rate-limit
description Bandwidth Control
match ip dscp 0
end
conf t
policy-map 3meg
class rate-limit
police 3072000 576000 exceed-action drop
end
This example is for a 3Mbps rate limit with a burst that follows the formula of (rate-limit * 1.5)/8
Then apply the policy to an interface:
conf t
int f0/1
service-policy input 3meg
service-policy output 3meg
end
This example is matching based on a DSCP value of 0. This is the default for most of the traffic on a network (mainly Internet) and as such, serves as a good match. The bad thing is anything that does not have a DSCP value of 0 in the packet is not affected by this rate-limit. This can be good or bad. Phones, both call signaling and voice, are marked with values other than 0 which is good as they are not affected by this rate-limit. Some p2p programs change their DSCP value to something other than 0, and as such, will not be affected by this rate-limit. I sugest a multi-level to get what you are looking for. At the switch level, things are handled in hardware. I would try to get the most of it here at the access layer first. Then at the first layer 3 boundary, I would also have another rate-limit on the interface or sub-interface catching anything that slipped passed the first rate-limit. For more detail, cisco press has an excellent book:
End-to-End QoS Network Design
Hope this helps.
