Hi,
Instead of "rate-limiting" the bad traffic, prioritize the "good" traffic like VOIP. If your Internet pipe is 80Mbit let's say for example you are seeing 70 % of this used by unwanted "peer-to-peer" traffic, why would you care unless you have an additional 50 % of traffic like VOIP that now drops packets. You would only care if "good" traffic doesn't make it through. This can be accomplished by using QoS for the good traffic. You could create 2 classes of traffic, one for good and one for bad traffic. You prioritize the good traffic so it could take up 80 % of your bandwidth (20 % for p-2-p so it doesn't starve). Your Internet pipe might always be utilized at 100% but this way you insure that the good traffic gets priority.
The general approach would be: Let customers use peer-to-peer applications if the bandwidth is available. In the case of congestion on the link to the Inernet, prioritize the good traffic using QoS. This is how most service providers implement "traffic control".
You can "mark" the packets as they come into our network, handle them accordingly in your backbone and then use those "markings" DSCP-codepoints, Diffserv or 802.1p tags to prioritize traffic going to the Internet.
Product-wise: Most of the managed ProCurve switches support QoS and so does the 7000dl router series.
Hope this helps.
