cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1615
Views
30
Helpful
33
Replies

QOS & H.323 Trunk

mightyking
Level 6
Level 6

Hello All,

I have integrated a CCM 5.1 (Main Office) with an Avaya S8400 G3 (Branch Office)using H.323 trunk connected directly from CCM to Avaya PBX over the WAN (T1). Everything seems to work fine except the audio quality. The average round trip ping reply is btw 100-120 which is quiet acceptable. We are using Cisco 2800 series routers in both ends. There's no Voice Vlan configured in the branch office. I though QOS may help to improve the situation. How am I going to prioritize the Voice traffic over data with H.323 Trunk? Any idea/config sample would be appreciated.

Thaks,

33 Replies 33

Wow, that was a pretty deep QOS training class my friend.

So back to my case, as I am using direct H.323 trunk, is the following what I need to configure in both end routers?

H323:

access-list 121 permit tcp any host 1.1.1.1 range 1718 1720

access-list 121 permit tcp any host 1.1.1.1 range 11000 - 11999

class-map match-any voice-rtp-nbar

match protocol rtp audio

Thanks,

Yes you would do something like this:

class-map match-any voice-rtp-nbar

match protocol rtp audio

class-map match-any voice-signaling

match access-group 121

policy-map

class voice-rtp-nbar

priority percent 40

class voice-signaling

priority percent 5

That would give you a pretty basic voice QoS. Depending on your link bandwidth, how many phones etc, you would want to play with those values.

-nick

Thanks Nick,

As of monday, I am starting to integrate another site which is anotehr Avaya S8400 G3. The site is connected to the Main site through Frame relay (FR). At the branch site we have a Catalyst 3750 L3 capable which is connected to the provider FR router. Can I use the same QOS config in the Catalyst 3750 as well? If not, what would be the best way to implement QOS in the 3750 end?

Thanks,

On the 3750 I would just use Auto QoS. Generally unless you have a large switched network and you're moving a lot of traffic, switch QoS will not affect much.

But no, the 3750 will use switched MLS QoS, and not routed QoS.

-nick

Hi Nick

why you used LLQ for signalling while garnteed bandwidth should be enough and better for signalling as it is not sensitive as VOIP ?

thanks

A few things on this:

1) For simplicity's sake.

2) I tend to see a lot of implementations where voice is the only important data going over the link, especially for remote sites. In this case, it may be better to prioritize the voice traffic.

3) Depending on your other data, voice signaling may still be more important, and some of the signaling is more delay sensitive that you would expect. Example: Pressing transfer on a IP phone requires a number of messages to be passed, including to/from the gateway for a PSTN call, and the cumulative delay of these messages equals the delay on the IP phone when the user gets a response.

4) For troubleshooting voice problems, I prefer not to spare any chance

If you're in a situation where the difference between a bandwidth command and a priority command is making a significant difference, you most likely need a fatter uplink anyways.

-nick

Hello Nick,

I was going to implement the QOS solution that you provided below. In the Cisco side we have a 2821 (data) router and in the Avaya side we have a 2610 (data) router. The IOS in the 2610 is : IOS (tm) C2600 Software (C2600-IS-M), Version 12.1(11), RELEASE SOFTWARE (fc1). I was thinking of upgrading the IOS to the c2600-advipservicesk9-mz.124-15.T8 in order to get better flexibility in implementing the modular QOS. The bandwidth btw HQ and Branch office is 10M. I am planning to have 5 simultaneous H.323 calls over the WAN. Do you think I can use the exact config that you sent me or do I need to play a bit with the priority percentage?

The Bandwidth utilization is over 90% all the time. Do you think implementing QOS will help to provide better voice quality?

QOS Config

class-map match-any voice-rtp-nbar

match protocol rtp audio

class-map match-any voice-signaling

match access-group 121

policy-map

class voice-rtp-nbar

priority percent 40

class voice-signaling

priority percent 5

Thanks,

MK

Hi MK,

Hopefully with a 10M connection you won't need much QoS at all. QoS should only really kick in when there is congestion, and you'll need a lot of traffic to fill up a 10 Mb pipe.

If you do G.711 you'll want to allocate around 80kb/sec per call.

For 5 calls this is 200 Kb/sec.

Thus, you would change your 'priority percent 40' to 'priority 200'.

You can also change your signaling to 'priority 15'.

For that few phones you shouldn't need to allocate 4 MB/sec, etc.

hth,

nick

I am sorry nick, I made a big mistake here. The bandwidth is a T1 and not 10M. Sorry for the mistake. Does it change the config a lot?

Do you recommand the upgrade to the new IOS version that I mentionned above?

Thanks,

MK

Hi MK,

I highly suggest moving from 12.1 code to 12.4 code. The commands we're discussing don't even exist in that code.

I would keep the same priority commands followed by the kbps.

If you want some more exact values, you can use the bandwidth calculator:

http://tools.cisco.com/Support/VBC/do/CodecCalc1.do

hth,

nick

Thanks Nick,

Let's say we have only one VLAN and all data and Voice are in the same VLAN. If a packet is not classified and marked in the switch, how a router knows which packet is voice and which one is data. My question may sound stupid but the reason that I am asking this is because I don't have access to the switches in the Branch office (more than 500 switches). All I have is access to the router. Would that be possible to priorotize a packet only in layer 3 without doing any layer 2 classification and marking. If so, how the router knows which packet is voice and which is data?

Thanks,

Hi MK,

If you look at the post I linked in the first post I went over some of these things.

Basically if you can't trust that the DSCP markings are going to be maintained you can do one of two things:

1) Match on an ACL for UDP on the voice range

2) Match with NBAR 'match protocol' command.

The ACL runs the risk of matching non-voice traffic as nothing prevents other applications from using those ports (16384-32767). This is layer 3+4

NBAR takes up processor resources reading the packets, but will give you a more accurate matching. This is layer 4+7.

hth,

nick

Thanks Nick,

Here's my very last question;

I have two router is 2 different sites. One is point to point and the other one is MPLS. I guess the QOS configuration would be enough for the one with Point to point link but do we need to do anything in the provider router with the one with MPLS connection?

When we priorotize a voice packet over the data, would the packet follow the same order in the MPLS network to the destination?

Thanks,

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: