12-15-2005 06:50 AM - edited 03-03-2019 11:14 AM
Need recommendations on how best to implement quality voice over Frame/ATM. Do separate data PVC and voice PVC make sense?
12-15-2005 07:50 AM
Hi,
the main issues you have to deal with are queueing and LFI. The best approach for queueing is LLQ (Low Latency Queueing). You might want to try "auto qos voip" on the respective interfaces to get started.
LFI (Link Fragmentation and Interleaving) takes care about jitter on a small (less than 768 kBit/s) access link. If VoIP packets have to wait for 1500 Byte data packet to be sent then it might take longer than the recommended 15 ms. The solution is to fragment on Layer2 with with fragment sizes less giving less than 15 ms jitter.
In the ATM case a second PVC is perfect for achieving this goal. "Fragment size" is 53 Byte - an ATM cell. Two Frame-relay PVCs will not help on this problem, better to use FRF.12 on one PVC.
Hope this helps
Martin
P.S.: There is a quite good QoS class (QoS 2.1) which explaines all that in detail.
12-15-2005 09:53 AM
Martin,
Just to clarify;
a. The link will be 1 Mbps. Does LFI still help?
b. If ATM PVC then use two PVCs.
c. If Frame Relay then use one PVC with FRF.12 fragmentation.
In either b or c, QoS still needed to guarantee good voice quality. Do you agree?
Thanks
Mario
Thanks
Mario
12-17-2005 04:19 AM
Hi Mario,
a) probably not, 1500 Byte on 1 Mbps give 12 ms jitter and that should be no problem.
b) Two ATM PVCs would help against jitter and would not be needed - see a)
c) FRF.12 is the implementation of LFI in Frame Relay - not needed in your case - see a)
In any case you would need priority queueing for VoIP and CBWFQ for VoIP signaling (it doesn´t make sense to prioritize VoIP if the telephone never rings as you drop signaling ;-)
So yes I agree.
Regards
Martin
P.S.: please rate helpful posts.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide