Does anyone know if there is a cisco SRND or doc explaining best way to design/implement QoS between multiple sites over a MPLS network, that is using Multilink frame relay on customer side to make connection to telco?
Have 4 T1s bundled at main site and they have it set to frame relay traffic shape to 4 meg cir with fragment of 1600 bytes, but one remote site has only 1 T1 and is a straight frame relay config (no MFR) and FRTS set to 1 meg so it doesn't match head end. Seeing voice quality issues at this site occassionally and not showing drops on QoS policies.
Trying to help someone out and reviewing an existing setup.
with FR you need to set the fragment size to a value matching the smallest line speed. Otherwise you might encounter jitter problems (voip packets waiting for max size data packets). With 1 Mbps and a maximum jitter of lets say 15 ms on the access line this would calculate to 1000000x0.015 = 1500 Bytes.
In addition apply a nested policy to the main site shaping the traffic for this site down to 1 Mbps. Otherwise you might have an overload situation at the remote site (ISP interface) leading to delays and discards.
Also make sure to limit the number of VoIP calls to the size of the priority queue (call admission control).
One more question: does the ISP offer QoS for your VoIP? The best would be to involve the SP.
If you have controcted for QoS COS then also make sure to match your LLQ to the appropriate values, as well as create a policer on the egress to police your Real Time Class as providers usually drop all the excess traffic in this class in MPLS networks.
Thanks for the reply and information. Good idea about a nested policy to shape it per site. They are contracted for QoS at the MPLS provider and I am waiting to find out what the specifics are of what they have signed up for.
The funny thing is the calls to Unity AA are only choppy/garbled that come in a PRI at this 1 remote site. G729 over WAN. A IP Phone call at that site to Unity or a IP Phone at the main site (also G729) works fine. The data and PRI gateway are the same 2811 router. MGCP. I think I found a possible problem on the unity server with NIC setup I am investigating further tomorrow.
Cisco usually recommends 33% for voip and 25% for best effort. This may not fall in line with the provider's Class of Service as my provider is giving COS of 40% voip and 6% best effort with the correct matching dscp values (ie voip = 40/46 and 24/26)
I am also in the process of setting up the QoS and have some intermittent Unity problems at the main site from the remote. The only difference with yours is I am on 711 for WAN but just curious on how you plan to fix the Unity? what I check so far is
1. If FA or E, Nic should be hard code but if Gig, then it should be auto which seems to be implied on latest revised CM 4.x srnd
Yea, the SRND is kind of helpful but I wish it was updated or I could find another document on cisco that had a sample design/scenario of a multi-site meshed MPLS network using frame relay on CPE side with single PVC to point to as a reference.
It's not just unity, I am just getting involved helping some folks out on this one and after gathering info, doing test calls, etc it is breaking up on other calls across WAN so is general WAN QoS issues/design. Just trying to figure out best way to tackle this since it's MPLS, meshed, and multi-site.
I'm not able to access my old voice mail messages all of a sudden. The recording says something like 'the message is currently not available'. This has never happened before in all the years I have been using this system. I have t...