Well..., no.
First of all, the bandwidth used is a function of the codec and protocol involved. G.711 is nominally 64 kbps, and with overhead it works out to about 88 kbps. High compression codecs will consume substantially less.
Second, the bandwidth used is NOT 64/88 kbps "in each direction" for an end-point. For a switch/gateway, where both end-points of a call traverse the pipe, your 2x math would apply.
It would be 64k (or whatever) in each direction.
Ivor
Ignoring, for the moment, silence suppression, you are correct, and the bandwidth figure for any given codec is the total for both sides of the conversation. The nominal bandwidth for a G.711 call is 64 kbps, not 128 kbps.
I was using G.711 as an example.
Please note the word "nominal", in the text you quoted. You are quite right that there is additional packet overhead that will drive the actual utilization of a G.711 closer to 90 kbps. The point I was trying to make for the OP was that it is that the nominal 64 kbps figure (or whatever is required for a given codec) is for the entire conversation.
Sorry, you've lost me. Speech in both directions passes down the cable from my telephone to the ATA and through my router to the outside world, so surely there is the same amount of data in both directions..?
Ivor
When 2 people initiate a voip conversation. Is a 64k channel for voice created both ways. In a 256K link would 128k be consumed?
Can G.711 even use silence suppression?
What about the packet overhead for the RTP header, UDP header, IP header and the ATM header? Isn't that going to add another 50% to the 64kbits/sec figure?
-wolfgang
For G711 each sampled packet is 64K + approx 12K of overhead and this is each way.. SO a conversation using full duplex will require approx
80K in each direction G729 is 8K + 12k therefore 20K in each direction. If your router supports stats you will see this.Ian
A router WOULD see such numbers, as it is handling TWO connections, one upstream and one downstream. An END POINT (soft phone, IP Phone, etc.) requires approximately 88 kbps for a G.711 conversation, not 196 kbps.
You can read here, for a more thorough explanation:
And in the real world a call is made up of two parts:- Upstream and downstream. and for example my sftphone is full duplex and I CAN see the conversation and the bandwidth is 79K in EACH direction. The site you mention is only taling about transmiting packets. Try it for your self. Setup calls on an ADSL line and your callers will get serious degridation after thr 3 or 4th call ( based on 256K upstream link wereas you wont hear any degridation.
This experiance is based on Lab testing and Packet sniffing with Mitel, Cisco and Asterisk IP PBXs
When designing IP networks the bandwith consideration has to be based on FULL duplex IE 80K or 20K in each direction. The fact that yes people often dont talk when someoe is taking to them doesn mean they wont. Or for even more of a laugh and to see the effects of your theory set the switch port to half duplex then try having a converation.
Ian
I little something I threw together, and always round up.
EBW=Estimated BandWidth in Kbps SR=Sample Rate in ms IPH=IP Header in Bytes L2H=Layer 2 Header in Bytes CD=Codec (ie G.729=8 and G.711=64) in Kbps
EBW=(((SR*(CD/8)+(IPH+L2H))*8)*(1000/SR))/1000
This in non-product related, it doesn't mater if it's Cisco, Nortel, Avaya, Vonage, Asteriks or anything else (despite what their marketing says).
L2H varies from as little as 6 Bytes to 38 Bytes and maybe more. It also will change depending on what you layer 2 is. It's ethernet at you PC but it's likely PPP at the router, use the highest number.
Tunneling adds another round of overhead and this isn't just VPN. PPoE, which is very common on residential DSL, is also a tunnel technology. Whith these the EBW becomes the payload of the VPN or PPoE packet.
Now this is just the start, we still have to deal with QoS, Serialization delay, Jitter, and Latency.
Have something to add? Share your thoughts — no account required.
Ask the community — no account required