QoS over Point-to-Point T1

[James Bond]

I'm fairly new to QoS, so I have a few questions.

I'm trying to prioritize certain traffic which goes over a point-to-point T1. I would like to give VoIP traffic high priority.

One end terminates with a 7204, and the other end terminates with a 2821. Would both of these be suitable to handle the task?

Anything I should know ahead of time?

 

[jlazzaro]

both are more than capable...is your voip traffic being classified and marked before arriving at these routers? bearer traffic should be marked with DSCP EF, signaling with AF31.

put all bearer traffic into a strict priority LLQ with guaranteed bandwidth to accomodate the voice traffic. control traffic should be in a non priority queue with a minimum bandwidth to ensure no packet loss.

from Cisco...

"LLQ is a feature that provides a strict PQ to Class-Based Weighted Fair Queuing (CBWFQ). LLQ enables a single strict PQ within CBWFQ at the class level. With LLQ, delay-sensitive data (in the PQ) is dequeued and sent first. In a VoIP with LLQ implementation, voice traffic is placed in the strict PQ.

The PQ is policed to ensure that the fair queues are not starved of bandwidth. When you configure the PQ, you specify in Kbps the maximum amount of bandwidth available to the PQ. When the interface is congested, the PQ is serviced until the load reaches the configured Kbps value in the priority statement. Excess traffic is then dropped to avoid the problem with Cisco's legacy priority-group feature of starving the lower priority queues."

 

[m1ldslide1]

What he said.

If your traffic is being marked close to the source (voip phone or gateway) and the markings are being trusted all the way to these devices, your config will be something like:

7204(config)#class-map match-all VOIP-payload
7204(config-cmap)#match dscp ef

7204(config)#class-map match-all VOIP-signaling
7204(config-cmap)#match dscp af31

7204(config)#policy-map VOIP
7204(config-pmap)#class-map VOIP-payload
7204(config-pmap-c)#priority [kbps assigned to priority queue]
7204(config-pmap)#class-map VOIP-signaling
7204(config-pmap-c)#bandwidth [kbps assigned to signaling]

7204(config)#interface s1/0/1:0
7204(config-if)#service-policy out VOIP

This config would provide LLQ to your voice bearer traffic and assure throughput to your signaling traffic. The real engineering comes in here - you need to know what the bandwidth requirements are for both. You'll need to know which VOIP codec you're using, packetization period, and other overhead (like cRTP vs. RTP, tunneling, etc) and find out what each call is going to require. Then multiply that by the number of simultaneous active calls you expect to have.

I typed all of this out so that you could get an idea of what all is involved and decide if it's within your skill level to take on at this point. You can always contact Cisco for help too.

 

[robmurphy]

If you are using G711 you will find that VoIP will use more bandwidth than sending speech down the T1 link as TDM traffic. With VoIP you need to add the RTP/UDP/IP headers. At 10 msec periodicity this will add 50% to the bandwidth needed.

I'm UK based so used to E1 not T1 but if my memory serves me connect a T1 = 24 * 64 kbits, of which 22 can cary speach. Using it for a data link I assume you use 23 channels, TS0 is kept for framing ect.

G711 speech at 10msec, over RTP/UDP/IP = 12 KBytes/sec
G711 over a timeslot, i.e. TDM, = 8 KBytes/sec

Using VoIP this gives 15 channels for speach, and leaves 4 Kbytes for signalling. Using the T1 in TDM would give 22 channels for speech, and a timeslot dedicated for the signalling.

Rob

 

[m1ldslide1]

Originally posted by: robmurphy
If you are using G711 you will find that VoIP will use more bandwidth than sending speech down the T1 link as TDM traffic. With VoIP you need to add the RTP/UDP/IP headers. At 10 msec periodicity this will add 50% to the bandwidth needed.

I'm UK based so used to E1 not T1 but if my memory serves me connect a T1 = 24 * 64 kbits, of which 22 can cary speach. Using it for a data link I assume you use 23 channels, TS0 is kept for framing ect.

G711 speech at 10msec, over RTP/UDP/IP = 12 KBytes/sec
G711 over a timeslot, i.e. TDM, = 8 KBytes/sec

Using VoIP this gives 15 channels for speach, and leaves 4 Kbytes for signalling. Using the T1 in TDM would give 22 channels for speech, and a timeslot dedicated for the signalling.

Rob

Wow so somebody actually uses those E1 circuits I keep having to read about?

G.711 produces 64,000 bps, or 80B per 10ms, and 160B per 20ms (typical voip packet).

So if his codec is g.711, he's looking at:

160B VOIP payload
40B IP/UDP/RTP header (uncompressed)
6-8B depending on PPP or HDLC (iirc)
= 208B per packet, at a packetization rate of 50pps requires around 83 kbps required per call. With cRTP, this is reduced to about 68 kbps per call. Like Rob said, this is more than a TDM voice call, which will be 64,000 (for the T1 you get 23 calls, 1 channel for CCS)

With g.729 on the other hand: 8,000 bps or 20B per 20ms, which without compression equals 27 kbps per call. With compression it's reduced to 12 kbps per call.

Obviously there are very significant OH savings in using g.729 or g.728 over g.711, but I'm guessing the OP doesn't have control over that. If the site is linked via T1 I'm guessing (hoping) that there aren't more than 30 or so users, so you can reasonably oversubscribe that T1 for these guys' VOIP so long as you have effective LLQ configured. Which I guess is the purpose of the whole thread...