VOIP Bandwidth problem

[goku-sun]

Hi all,

I have questions for you VOIP experts...

Are there much issues for a point-to-point VOIP service between 2 locations if ping delay between 2 ends is as large as 580ms (Routers=cisco 1750, meduium is Satellite)

Suscribed bandwidth is 128k but the voice quality is poor and often only one side can hear but the other can't.

TIA,

Goku

[TIMOTHY DAVIDSON]

In my experience, VoIP doesn't work well on anything more than about 180ms of latency. 580ms is way beyond the recommendations from Cisco. Most wireless technologies, especially satelite, introduce too much latency to work well for VoIP, which is why Cisco won't currently support them. They also suck for games. :-)

As far as bandwidth is concerned, as long as you use QoS and do your calculations correctly (to determine the number of concurrent phone calls), quality is great.

[jwang]

580ms is Okay. but 128K is too low for calls without cRTP. if it's p-to-p link and serial interface on router, you should use compressed RTP and probably G.723.

[carlos]

You really need QoS for this setup. I´ve done it with a 64Kbps satellite link. Ping response is 600ms average. Got two simultaneous voice channels working with acceptable quality.

The only issue is the call setup, it takes about 10 seconds to establish the call. Oh.. and data will suffer because of the low MTU.

Here´s the relevant configuration

SITE A

Cisco 1750 2 FXO ports connected to a NEC PBX.

----------------------------------------------

interface Serial0

description Enlace a Laredo

bandwidth 64

ip address 10.1.1.1 255.255.255.252

ip mtu 150

encapsulation ppp

ip tcp header-compression iphc-format

no keepalive

fair-queue 64 16 2

traffic-shape rate 64000 8000 8000 1000

ip rtp header-compression iphc-format

ip rsvp bandwidth 48 24

!

!

voice-port 1/0

timeouts ringing 20

timeouts wait-release 3

connection plar opx 200

codec g729r8

!

voice-port 1/1

timeouts ringing 20

timeouts wait-release 3

connection plar opx 201

codec g729r8

!

dial-peer voice 101 pots

destination-pattern 5

port 1/0

!

dial-peer voice 102 pots

destination-pattern 5

port 1/1

!

dial-peer voice 301 voip

destination-pattern 20.

session target ipv4:10.1.1.2

dtmf-relay h245-alphanumeric

req-qos controlled-load

!

!

SITE B

CISCO 1750 2 FXS ports with ordinary phones attached.

-------------------------------------------------

interface Serial0

description Enlace a Cd. de Mexico

bandwidth 64

ip address 10.1.1.2 255.255.255.252

ip mtu 150

encapsulation ppp

ip tcp header-compression

no keepalive

fair-queue 64 16 2

traffic-shape rate 64000 8000 8000 1000

ip rsvp bandwidth 48 24

!

!

voice-port 1/0

!

voice-port 1/1

!

dial-peer voice 101 pots

destination-pattern 200

port 1/0

!

dial-peer voice 102 pots

destination-pattern 201

port 1/1

!

dial-peer voice 301 voip

destination-pattern 5

session target ipv4:10.1.1.1

dtmf-relay h245-alphanumeric

req-qos controlled-load

!

!

Hope this helps.

Carlos

[goku-sun]

Many thks Carlos,

I indeed appreciate your experience sharing.. May I ask if the access routers connecting your 1750's need any special qos config? Are those commands also involve compression apart of BW reservation?

My existing config uses frame-relay dividing the serail into 2 PVC's(one for voice and the other for "intranet") by traffic shaping commands. I tried to dedicate the whole 128k BW for voice but still one of the conversation(only one voice call active) will die after a miniute or so. Therefore I am right in guessing there may be other issues around..

cheers,

Goku