RTP traffic through a SIP Trunk between Avaya and CUCM ?

[joby.joseph]

Hi Voice Experts?

I have 4 sites connected through MPLS cloud and using Cisco CUCM 6.1 based voip solution for the internal calls.The  bandwidth is with a premium CoS and dediccated for Voice.

1, Location P - 1 Mbps bandwidth, hosting Cisco CUCM publisher(7825H3), Subscriber (7825H3)and Unity Connection (7835H2), Voice gateway & Hardware Conf Bridge (Rtr - 2811 with 4DSP).

2. Location R - 3 Mbps bandwidth, having 200 IP phones (Cisco 7940), Voice gateway & Hardware Conf Bridge (Rtr - 3825 with 8 DSP)

3. Location B - 1 Mbps bandwidth, having 100 IP phones (Cisco 7940).

4. Location K -  3 Mbps bandwidth, hosting Avaya S8500 & G650 based voice solution.(Another company)

For the calls between my company and the other company, we have SIP Trunk configured between Cisco CUCM and Avaya Proxy. We are observing quality issues in voice calls between 2 companies. After analyzing the traffic , we have following observations.

1. The bandwidth utilization in Location P ,where CUCM hosted ,is high. And the traffic is busty in nature and some drops are observed on priority queue.

2. The calls between companies uses G.711 codec .

3. The SIP Trunk is configured with 'Media Termination Point (MTP) Required settings.

For the calls from locations R & B to location K through SIP trunk, all the RTP traffic is passing through Location P in addition to signalling traffic. Is this behaviour is normal? I was thinking that only signalling traffic with flow through CUCM and RTP traffic will be between endpoints.

In our case MTP media resource is configured on Publisher & Subscriber.. Is this the reason that RTP flow to location P?  I want to reduce some traffic on location P. So if I configure a router with DSP in Location R as MTP media resouce, and use that MRGL for the trunk, will it help for offloading the RTP traffic (through SIP trunk) from location P?

Thanks

Joby

Appreciate if you can share your comments and valuable suggesion on this.

[Steven Holl · Accepted solution]

1. The bandwidth utilization in Location P ,where CUCM hosted ,is high. And the traffic is busty in nature and some drops are observed on priority queue.

This is the result of sending too much traffic into your prioriy queue.  You either need to reduce the traffic through this queue, or increase the size of the prioirty queue or burst interval until drops stop occurring.  If this is MPLS, your real-time queue size that the provider has provisioned should match what the router is provisioned for.

2. The calls between companies uses G.711 codec .

If you switch this to g.729 you'll cut your bandwidth on the trunk down by ~75%, which will help with the above point.  Note that if you hit g711-only applications across the WAN (Unity, AA, IPCC, etc) then you will need a transcoder at the site where the g711 application sits.

3. The SIP Trunk is configured with 'Media Termination Point (MTP) Required settings.

Do you have a reason for this?  It isn't necessary in all designs.  If you switch to g729 as mentioned above, and need to keep an MTP, you need to configure a g729 MTP which would be a software IOS mtp.

For the calls from locations R & B to location K through SIP trunk, 
all the RTP traffic is passing through Location P in addition to 
signalling traffic. Is this behaviour is normal? I was thinking that 
only signalling traffic with flow through CUCM and RTP traffic will be 
between endpoints.

This is because you required an MTP, and the MTP chosen is likely at site P.  Move your MTP to the site of the SIP trunk to prevent media from flowing to an unrelated site.

You can do all of this without DSPs (unless you need transcoders).  IOS Software MTP in g729 in the way to go, in my opinion.  When creating the MRGL, make a MRG JUST with software MTPs, and put it in the top.  Then have your other resources (xcoder, cfb, etc.) in a separate MRG below the IOS SW MTP MRG in the MRGL assigned to the device.

-Steve

[saheed.kv]

Hello,

Is it possible to configure IOS Software MTP in MGCP gateway, if yes how it is done? Does a call comming from PSTN > MGCP gateway (G711u) to SIP trunk or SIP phones require MTP or Xcoder at any point of time?

//Saheed

[Steven Holl]

saheed.kv@volvo.com
Hello,
Is it possible to configure IOS Software MTP in MGCP gateway, if yes how it is done? Does a call comming from PSTN > MGCP gateway (G711u) to SIP trunk or SIP phones require MTP or Xcoder at any point of time?
//Saheed

Yes.  The procedure is the same for any IOS SW MTP, regardless of the GW protocol.  GW protocol is independent, since the MTP is registering to CM, not to the gateway.  The config looks just like it would for a transcoder, other than the profile being of type MTP with 'max sess software #', and added in CM as an IOS Enhanced MTP.

That call flow shouldn't need a transcoder if all sides support 711.  Though it will invoke an MTP if you have a DTMF mismatch.  Configure rtp-nte on both the MGCP gateway and on the SIP phone/trunk to prevent that.

[saheed.kv]

Many thanks Steven Holl,

This is the kind of answer I was looking for.

BR//Saheed

[joby.joseph]

Hi Steve,

Thanks a lot for your detailed response.

1. As you mentioned,the excessive traffic is due to the usage of G.711 codec. If the number of call in SIP Trunk is high, the circuit becomes busty and overutilized. Yesterday we have an incident. During peak time, there was nearly 10 calls on SIP trunk and many phones lost connectivity to CUCM and shown switching to SRST mode. There was nearly 20% of drops observed on the MPLS ckt (Site P - 1 Mbps)  at that time.

Is it possible to limit the concurrent calls through the SIP trunk? . We are using static location based Call Admission Control and the SIP Trunk is also part of a common location named DC-LOC. If I define a new Location with 800 kpbs bandwidth and move the SIP trunk to that location, will it help for call admission control ?  (Location R is configured with 2400 and Location B is configured with 1024)

2. I am not sure why the MTP Requirement was enforced on the trunk. This setup was done 1 year back and used from then. Since the other end is Avaya SIP system, I think it is done for some DTMF conversion or SIP Early offer support with CUCM.

Also the Avaya systems (the other end of SIP trunk) in the other company is in a DMZ setup. So for them it will be easier to manage traffic from few set of IPs rather than allowing the entire voice VLAN subnet from our company.

I configured a test MTP resource in location R (Router 2851 with 3 DSP) with the support for 15 hardware sessions and 15 software sessions. But after using this MTP, I couldnot make calls to other company. My phone shows the call is connected but there was no audio.. Seems the RTP traffic from new MTP is blocked in DMZ firewall.I have initiated a change to allow the traffic from new MTP source before further testing.

Here, I have a doubt.. What should be MRG and MRGL for the SIP Trunk ? Should we make both the endpoints  (sites R and B) and SIP trunk (site P) in the same MRG /MRGL for making use of new MTP  or only endpoints should be moved to new MTP configuration?

Thanks and Regards

Joby

[Steven Holl · Accepted solution]

Is it possible to limit the concurrent calls through the SIP trunk? . We
 are using static location based Call Admission Control and the SIP 
Trunk is also part of a common location named DC-LOC. If I define a new 
Location with 800 kpbs bandwidth and move the SIP trunk to that 
location, will it help for call admission control ?  (Location R is 
configured with 2400 and Location B is configured with 1024)

Locations are probably the best way to do that.

2. I am 
not sure why the MTP Requirement was enforced on the trunk. This setup 
was done 1 year back and used from then. Since the other end is Avaya 
SIP system, I think it is done for some DTMF conversion or SIP Early 
offer support with CUCM.

You only need an MTP for DTMF when you need to convert DTMF from DTMF in the media stream to DTMF in the signaling path.  So it's only needed if one side doesn't support rtp-nte, or of in-band voice is negotiated for DTMF on one side.

I 
configured a test MTP resource in location R (Router 2851 with 3 DSP) 
with the support for 15 hardware sessions and 15 software sessions. But 
after using this MTP, I couldnot make calls to other company. My phone 
shows the call is connected but there was no audio.. Seems the RTP 
traffic from new MTP is blocked in DMZ firewall.I have initiated a 
change to allow the traffic from new MTP source before further testing.

Sounds like a routing or L4 issue.  Make sure the interface you are using for the MTP (sccp local interface...) is routeable to both endpoints.

Here, I have a doubt.. What should be MRG and MRGL for the SIP Trunk ? 
Should we make both the endpoints  (sites R and B) and SIP trunk (site 
P) in the same MRG /MRGL for making use of new MTP  or only endpoints 
should be moved to new MTP configuration?

For your environment, I think getting local MTPs for the devices is the best idea.  Use the appropriate MRGs such that you can take advantage of a resource local to the device requesting it wherever possible.