Reasonable amount of WAN head-end over subscription

rberke · ‎11-29-2007

We have an E1 Frame Relay connection at our head-end site, with 2 remote sites at 512K each. We have been using a sub-interface with one PVC, and with WFQ and FRTS applied per remote site. The overall serial link has had no WFQ; just FIFO. We've shaped the outbound traffic low enough to prevent exceeding the remote site's 512K port.

We're upgrading the remote site ports to E1, and want to send out a higher rate from the head-end to them. What's a reasonable amount of oversubscription of the head-end E1 (the sum of FRTS of the remotes) to configure? 1.5x? 2x?

Would you recommend simply applying WRED to the overall serial interface, to try to reduce the impact of occasional heavy demand towards both remote sites at the same time? Are there restrictions against that, with the subs defined?

Thanks,

Richard Berke

T. Rowe Price

smalkeric · ‎12-06-2007

If you have over-subscribed you should consider traffic shaping as a possible solution.

http://www.cisco.com/warp/public/732/Tech/gts/

http://www.cisco.com/warp/public/125/framerelay_ts_cmd.html

rberke · ‎12-07-2007

We have Frame Relay traffic shaping configured on each of the sub-interfaces. It's their sum that I'm wrestling with on the 'parent' interface.

I didn't think you could apply a 2nd layer of 'fancy queueing' for the whole interface.

Richard