Discussion Locked

Traffic Engineering

0 helpful
6 replies
Closed to new repliesThis discussion is closed. It stays readable, but it takes no new replies.

Hi,

Currently, I have 3 new ITSP customers and they are planning to upgrade
their backbone network. At the moment, they have 3 E1 lines to Local
ISPs in Malaysia and planning to have 2 E1 connections to UUnet soon.
Anyway, is there any formulaor rules that we engineer to work on in
order to decide how much bandwidth required for VoIP business? For
example, if I am planning to have 1 million subscribers and 500,000
users will be active at the time, then how much bandwidth do I require
in order to offer good voice services? Anybody any idea? A lot of local
engineers keep on telling me that we can only guess for the bandwidth
and there is no exact answer? Is that true? I really doubt. Thanks for
your advice

6 replies

by fnugry

Most of the bandwidth consumption comes from the call traffic, which highly depends on the voice codecs you use, as well as on codec setup (normally you have a number of quality options to choose from). Common standards are: G.711, G.729, G.729a, G.726, G.728, and G.723.1. Below is a list of common data rates for standard codecs - those can be used to estimate bandwidth consumption when you know high/low/avg number of simultaneous calls you want to support. The problem, however, is estimation of those highs/lows - here you're on your own. The only thing I can suggest is get number of customers who will use VoIP from your business plan and use statistics from local telco to estimate number of simultaneous calls in peak/off hours for that number of users. Additionally, you should coordinate bandwidth purchases with purchase of gateways, switches, servers, PSTN lines, etc (there is little sense in buying bandwidth enough for 100k calls if your server/gateway can only suppot 1k). I mentioned that call traffic is the major part, additional traffic comes from call setup - and this depends on the VoIP system you use. See H.323 spec for some estimations.

G.711 A-Law 64,000 bps
G.711 u-Law 64,000 bps
G.723.1 5,300 bps
G.723.1 6,300bps
G.726 16,000 bps
G.726 24,000 bps
G.726 32,000 bps
G.729 ANNEX-A and B 8,000 bps
G.729 ANNEX-A 8,000 bps
G.729 8,000 bps
G.726 24,000 bps
G.726 32,000 bps
G.728 16,000 bps
G.729 ANNEX-A and B 8,000 bps
G.729 ANNEX-A 8,000 bps
G.729 8,000 bps
G.726 24,000 bps
G.726 32,000 bps
G.728 16,000 bps
G.729 ANNEX-A and B 8,000 bps
G.729 ANNEX-A 8,000 bps
G.729 8,000 bps

by csmith

I think the engineers are also concerned about the "jitter" problem.
This is caused by the IP packets arriving with different delays.
Multiple line aggrevate this problem, in some instances, and bandwidth alleviates it. To a degree.
The engineerscan make an "informed guess", but absolute accuracy is not possible, because of this problem.
The other side is, a reasoned guess (estimate) is usually good enough.
Regards, Chris

by csmith

,
Example:
If it is assummed G.729 with Compressed RTP Headers, the bandwidth required for 500,000 people making IP telephony calls simultaneously is 6.4 Gbs.
Note: That is Giga not Mega.
Do you see what they are talking about?
Regards, Chris

This discussion is closed to new replies.