VOIP Branch Call Quality Issues

[mklein76]

We have CS1000S at the main office, full voip, CS1000B's at branch offices and symposium 5. Branches are connected to the main office via MPLS network with QoS and calls travel from the main office where Symposium resides to the branches over virtual trunks.

We get intermittent complaints from users at the branches that incoming symposium calls are garbled, the volume goes up and down and sometimes they can't hear the first 5-10 seconds of the call. We've done troubleshooting to try and isolate the problem, patched up media cards, swapped out media cards at the branches. The problem is that it's pretty hard to replicate and it only happens to a few specific people.

I'm starting to think it might be a QoS issue but I don't handle QoS on our MPLS network and the guy who does run it, says it's working correctly. We are looking into getting PVQN/Vivinet but at $100 per phone/device to monitor, that's a little steep. Is there anything else we should be doing to try and isolate the problem?

Matt
Forex Capital Markets
CS 1000E VoIP/Symposium Admin

 

[jbrown1217]

Before you had the VoIP system installed was there a network assesment done between all locations? Just because someone says that it is working correctly does not mean that it is. If the problem does no occur on a consistent level and all of the media cards, signaling servers and call servers are up to date with patch levels and firmware I would look at testing the network itself. Your voce quality is only going to be as good as the network that it is being transmitted across.

 

[mklein76]

Well, we have a 45 mbit DS3 connection to the branch and we never fill it up so I'm told by the networking guy that QoS shouldn't be an issue. Since I don't do cisco and QoS, it's hard for me to tell him what to do or what to look for. Before we had the nortel phones, we had tadiran ip phones connected to a point to point T1 and we had no issues with call quality like now so I'm pretty confident in our infrastructure with the exception of how QoS is working. Are there any other products out there that will trace a call thru the system and diagnose problems?

Matt
Forex Capital Markets
CS 1000E VoIP/Symposium Admin

 

[jbrown1217]

Is voice being placed into a priority queue before entering the MPLS and is that priority the same when it reaches the branch? Are you having any call quality issues when calling station to station at the branch or the main site? If you have access to the Nortel ntp's (553-3001-160,553-3001-365, and 553-3001-214)there are explanations in there about QoS and Nortel's requirements.

 

[dlesap]

believe jbrown...
network assessment is a must, doesn't matter what network guys says. The bandwith is not an insurance of the voice quality. Only trust the result of the assessment. I recommend to use NetIQ which is a purely tool for voip. Reports are clear and understandable.

 

[teekow]

Start with confirming the class-map and policy-maps that are assigned to the outputs of the WAN interfaces on the routers. Make sure that the LAN/WAN guy knows you are using DSCP 40 for signaling and DSCP 46 for RTP. Cisco defaults are different. You should have 2 different class-maps and 2 different policy-maps for the DSCP values above. Signaling needs a bandwidth statement and RTP needs a priority statement which matches what you setup in the zone bandwidth. Remember that the priority statement in the router is full-duplex and the zone bandwidth value in the Nortel node is half-duplex. Hence, if your priority queue is 4000kb, your interzone bandwidth should be 8000kb. That assures you do not send more calls out on the WAN than the QOS is capable of. After that talk to him about trusting DSCP on the Cisco switches. By default, the switch ports are not trusted and the dscp value is discarded (this goes for Nortel POE switches as well). After that, you can have him use the "Show Policy-Map Interface" command on the routers and you can see the packets ending up in the right queue on the routers.

Good Luck
Rob