Our internet pipe is getting pounded (receive/return traffic) since we upgraded our school's T1 to Gig. My question, would the only way to improve the congestion be with a bandwidth upgrade? I can't perform policing or shaping on received traffic from the isp on our router's outside int right?
If I did do a service policy to police traffic at lets say 256K, wouldn't the clients adjust the tcp window accordingly, which may help?
Any suggestions or comments are welcomed. Thanks in advance
Your reference does discuss the 6500 series Microflow policing feature I had in mind.
As for other documentation, don't have any good references because what we're dicussing is a bit on the extreme edge of QoS. Most documentation would assume you have the capability to manage bandwidth upstream of a bottleneck, not downstream of it. Most QoS documentation really doesn't address TCP per-flow bandwidth management by drop management (except perhaps RED). We're also limited by the capabilities of typical network devices. Applicances (e.g. Packeteer's) designed for bandwidth management (which I don't think Cisco has within their product line), provide other interesting features. For instance, spoofing a receiver's TCP RWIN can be used to regulate a TCP sender's transmission rate.