Multiple PVCs sharing the same TX ring?

[jasosmith]

In an implementation where there are multiple PVCs on the same Frame-Relay circuit, how does the TX ring service each PVC?

I have been able to find good documentation as to how packets are put into either the TX ring or the Output Hold Queue ("Inside Cisco IOS Software Architecture" link at bottom of page), but none that addresses multiple PVCs on one interface.

The general setup:

The TX ring queue has been reduced to 4 to force packets into the output hold queues. Fragmentation is set appropriate to the BW. CBWFQ w/ LLQ is running on both (all) PVCs.

Assumptions are as follows:

1. Each PVC may have voice traffic destined to it.

2. There is one output hold queue per PVC. (Since there may be one instance of CBWFQ w/ LLQ running on each indivicual PVC).

3. I'm assuming that there is one priority queue for each output hold queue. (The answer may be that there is only one priority queue for all output hold queues, but I don't believe that is the case).

Some or all of my assumptions may be incorrect, and thus the reason for this post.

Possible answers as I see them are as follows:

1. Straight, plain FIFO.

2. Some sort of WFQ where the TX ring will service queues based on their configured BW of the PVC.

3. Something I haven't thought of.

Thanks

Jason

Great reference.

http://www.amazon.com/exec/obidos/tg/detail/-/1578701813/ref=pd_sxp_f/002-6869153-1162428?v=glance&s=books

[beng]

Here are couple from cisco site:

http://cco/en/US/tech/tk39/tk824/technologies_tech_note09186a00800fbafc.shtml

[jasosmith]

Thanks for the info. The link you referenced is in regards to ATM so am I to assume that the same holds true from Frame, say a PA-T3?

From the site above, my understanding is that there is a seperate TX ring for each PVC? At least this sentence leads me to believe that... "In general, the size of the transmit ring is based on bandwidth of the interface or VC and is a power of two "

This would also lead me to believe that there is a singe output hold queue for each TX ring and therefore a priority queue for every ring/hold queue.

On the right track? If this is true however, how are does the PA determine which of the TX rings it services and how? I think I am still missing something.

[jasosmith]

Actually, answered my own question with your help...thank you very much for the info.

For anyone who is interested, here is the short synopsis with the link to follow.

Frame-relay does not support multiple TX-rings for each VC (as ATM does). It does however, support three different queuing mechanisms at the interface queue level. These are, FIFO, DUAL FIFO and PIPQ. In short, when you enable fragmentation, you are also enabling DUAL FIFO on the interface queue. With DUAL FIFO running, your high priority packets are put into the priority queue in DF and therefore processes accordingly. PIPQ is another possibility, and is designed primarily for when you are running seperate protocol types on different VCs. (for example you break out a PVC strictly for voice).

Here is the link.

http://www.cisco.com/en/US/tech/tk543/tk544/technologies_tech_note09186a00800a4754.shtml

Thanks all.

js