I have a situation I really need some advice in. Here's the scenario:
At our HQ, we have a 3640 with a couple hundred ATM IMA PVCs. Each of these PVCs get switched to individual Frame Relay PVCs in the telco cloud, and terminate as such at each of our stores with a class C. At one particular store, we split the class C into subnets. One specifically handles critical store operations, and another handles non-critical accounting stuff. We need to prioritize data across this PVC somehow, so that at any given moment the free bandwidth of the link can be used by any operation, but packets coming from the critical store operations need to take precedence.
In other words, if nothing is going on at the store, the accounting department needs to be able to max out the link with, for instance, a file transfer, but if someone hits a keystroke on one of the store terminals, we need for that data to immediately take precidence over all queued traffic from the file transfer in both directions (the store terminals require as low a latency as possible for each typed character to be echoed back to the screen by the remote server. Any delay will make things very frustrating for the personnel).
At first I thought we should implement priority queuing for this, but I'm not sure if this is possible on a per PVC basis. From what I've found, this appears to only configurable at the interface, and not the subinterface level. I can't have one priority situation on one PVC affect all the other PVCs on the same physical interface on our core router.
At this store we have a 128k fractional T1 Frame Relay circuit with a CIR of 64k. At our headquarters, this circuit switches to one of many PVCs on our ATM IMA group, which is carried over three full T1s. Any suggestions?
Thanks!
