SIP Trunk from Cisco CallManager to Service Provider

[nshoe18]

We have a SIP trunk from our CallManager 8.0 to our Service Provider and are having some QoS issues with jitter, call drops, etc. How to you assign QoS to a trunk that is terminated to CallManager and not to a router?

[tobin hawkshaw]

Call manager automatically marks its signalling and voice traffic with cs3 and ef respectively, you can change these settings in service params.

This is obviously for outbound traffic.

Inbound should already be marked by you SP.

[nshoe18]

So CallManager marks it but I am looking at what should be setup in the switching and router to facilitate proper QoS and call setup.

[tobin hawkshaw]

QoS is a fairly broad topic so if you dont have it configured already you have allot to do, Ill work on the assumption that the lan can do without it for now, can you give me more information on the problem.

- Are the bad calls only across the trunk

- Is it all calls or only when you have a few simultaneous calls

- What else is the link used for

- Whats the topology look like

[nshoe18]

Here is a short run down:

1) Calls are experiencing a lot of jitter when we are calling out or calls some in....the other end of the call does not hear the same thing

2) Seems to be when multiple calls and especially when downloads are occurring from the Internet, the SIP trunk is carried across a bonded T1 that supports Inet data as well

3) Topology is as follows:

CUCM -----Switch via gig link -------Main switch via gig link -------- 2821 router via gig link -------- Inet ---- SIP Provider

The switch that the CUCM is plugged into has AutoQoS setup for our phones on all the ports that support phones but I have not put anything on the CUCM port as I am unsure of what would be best. The CUCM is setup with the trunk to the SIP Provider not the router.

[tobin hawkshaw]

cool, setup QoS on your routers wan connection, I very much doubt you need to set anything up between call manager and the router with them being Gb connections, i would expect the problem is at the only choke point, the WAN interface.

you could run autoqos on your internet facing wan interface and then adjust the settings it puts in the policy map, set it to something you want to reserve for just voice traffic (usually around 33%), this should setup LLQ using the priority command under the policy map, it should then protect your VoIP traffic from the downloads etc...

You should probably use locations based CAC to stop too many calls using up this BW as well.

whats the internet connection like? is it pretty fast including the latency?

If you need help with any of the commands let me know.

[nshoe18]

OK I am trying to implement the QoS stuff and it appears that maybe the AutoQoS is not available on Gig interfaces? I may need those commands.

[tobin hawkshaw · Accepted solution]

no worries,

first we need to classify the traffic, Im going to do this the lazy way using nbar, but you can take the basic principle and edit it for your needs if you like.

ip cef (you probably have this enabled anyway but hey..)

class-map  VOIP-RTP

match protocol rtp audio

match protocol rtcp

class-map VOIP-SIG

match protocol SIP (if your using sip out, im guessing you are)

policy-map  VOIP

class VOIP-RTP

     priority (enter how much bandwidth you want to prioritise on the link, usually up to 33% of bw, you can also enter a percentage)

class VOIP-SIG

     bandwidth (usually around 5% of left BW, or with a percentage if done above)

class class-default

     fair

interface 0/1 (example)

     service-policy output VOIP

Test this after you apply it and see if it makes a difference.

please rate helpful posts

[nshoe18]

Thank you. I put that in and we are testing for the day and I will let you know.