 This topic has 4 replies, 1 voice, and was last updated 15 years, 9 months ago by Ed.

AuthorPosts

CCDGuest
Hi! i have a similar doubt about VoIP.Right now i’m doin’ a ‘project’ about this issue and im quite confused with the bandwith requirements as well.Using a G.711(64Kbps) or G.729(8Kbps) codec, which will be the new bandwith including RTP/UDP/IP headers?? 80Kbps per each direction in G711 case???in G729???.
If i have to offers full VoIP service to 648 customers in the rush hour at the same time, what kind of gateway do i have to buy??How do i know de number of customer it can give service to?how many calls can it support??…thank uWilson BoyrieGuestFor g729, about 20kb per active call.
To be absolutelly sure, calculate 22 to 23 kb of bandwith per active call on g729.
For a DS3 (672 channels)you will need a BIG switch.
Most regular gateways go up to a handfull of T1 ‘s only.
Above 10 T1, you are talking only a handfull of very expensive vendors.CCDGuestTHANKS! and what about G.726 codec? lot of devices im gonna work with require this codec, i think it works at 32Kbps without headers.. doesn´t it? how many bytes add RTP/UDP/IP/ethernet??
StivGuestIf you are using g.729 and sending this with few frames (up to 8) you could reduce BW to less then 13k.
My question is about practical comparation and experience with g.729 and g.723.1 (6.4k)
G.723.1 has MOS 3.7 and g.729 has MOS 3.9, but
g.723.1 is more robust and if you have noisy lines, low Voice level or higher PL, g.723.1 has better overall performances and use 12.7k of BW.EdGuestFirst you have to find out how your holding time is distributed. normally this should be negative exponentially. Then you must determine the average holding time. After that you can calculate the probability of x call arrivals during unit time assuming that the arrivals do not overlap and that the probability of x arrivals in time t equals.
This can be done using the Poisson formula.
After you have calculated the Poisson distribution, just sum up the probabilities of having n arrivals until you come to (for example) 0.99. Then the probability that (due to too many arrivals at the same time) the system utilization exceeds 100% is 1%. That means, if you have summed up the probabilities from 0 arrivals up to (for example) 25 arrivals, that the probability of upcoming arrivals not being served is 1% after already having 25 arrivals.You can also calculate the blocking probability using one of Erlang equations (dependet from the type of VoIP system you use).
To calculate the bandwidth, you must consider, that not all bandwidth can be used. (e.g. today’s switched Ethernet allows a maximum of 90% bandwidth utilization)

AuthorPosts