Archived from groups: comp.dcom.voice-over-ip,comp.dcom.sys.cisco,comp.dcom.sys.nortel (
More info?)
"Joe Technician" <Joe@somewhere.com> wrote in message
news:hinhc.63945$dg7.45876@edtnps84...
> Questions questions questions.
>
> The point to point T1 for Data would just be a 1.544M data stream. So the
> question arises, what compression (if any) would you use? No compression,
> no way you'll get 50 users on there.
you need to be careful what you are talking about here. The number of users
at the site may not be the same as the number of simultaneous calls you
support offsite.
e.g. if this is a call centre, then there should be 1 external voice "line"
per agent or more (or you cant run the call centre at full load, or cant Q
waiting calls etc).
if it is a business site, then there are normally fewer lines than users -
1:4 is a ratio sometimes used for offices at work, but it really depends on
how much phone use is likely, how many calls go outside, how near the worst
case peak you want to allow for.....
Also, the scenario the OP described doesnt mention where / when calls go out
to the PSTN - given 3 sites there could be a public voice connection at each
site, just at 1 site or another combination - the arrangement will change
how much of the total offsite voice traffic needs to go down each T1 link.
How much Data is on the link sharing
> bandwidth with the VoIP traffic, or is it a dedicated link for VoIP? If
> there's data, do you have QoS on it? Do you have Modem lines? Fax and
> Modem lines are more susceptible to compression then voice. What kind of
> jitter, and latency do you have on these T1s? How susceptible are they to
> packet loss?
And what is the voice encoding? if you look at the cisco design guides for
call manager, they assume G.711 for LAN calls, and G.729 for WAN calls. The
bandwidths would work out at 80k and 14 to 28k per call respectively (both
are full duplex)
have a look at the call manager design docs at
www.cisco.com/go/srnd
they will at least give you an impression of what such a design would look
like.
>
> So, in other words, there are so many possible variables to consider to
make
> VoIP work that you won't get them all in a note, and consultants charge
> around $5,000 per site just to evaluate it.
>
> JT
>
>
> "Walter Roberson" <roberson@ibd.nrc-cnrc.gc.ca> wrote in message
> news:c64ujj$4j2$1@canopus.cc.umanitoba.ca...
> > In article <ac9830d7.0404202025.50c499e5@posting.google.com>,
> > Trilok <trilok1@coolgoose.com> wrote:
> > : I am an IT graduate student and doing a term paper on Voip
> > :implementation.
> >
> > : 1.How many phone & fax lines can be accomodated on the T1 line?
> >
> > A channelized T1 would have 24 independant timeslots, 1.544 megabits
> > per second total. Each timeslot could be used to carry a telephone-
> > company quality call of 8000 samples per second, 8 bits per sample.
> > (The extra 8000 bits are used to carry control information.)
> >
> > That's for standard calls. VOIP would, though, typically use IP
> > and compression techniques to reduce the data stream requirements,
> > and for VOIP you wouldn't necessarily want to channelize your T1.
> >
> > : 2.Will the T1 line be able to handle all 50 users simultaneously?
> >
> > You can always adjust the VOIP lossy compression algorithm parameters
> > until the data fits. You need a minimum-quality metric in order to
> > make a decision about how many VOIP can be carried.
> >
> > : 3.How do you calculate the above(1) & (2)?
> >
> > (1) is by specification of T1, which you can research in IETF
> > standards (or just look up on some page or other at cisco.com)
> >
> > (2) is the much more difficult question, as it depends upon the
> > quality of your perceptual coding algorithms and upon your standards
> > of intelligability at the other end. It also depends on whether the
> > VOIP is truly being used to carry -voice-, or if sometimes you want
> > to run fax over it, or if sometimes you want high-quality music...
> >
> >
> > : 4.What is point-to-point T1 line: does it mean that there is a
> > : dedicated line that runs between one location's Lan to the
> > : other location's Lan?
> >
> > Yup, pretty much. It might go through some switching equipment at
> > various telco's along the way, but there would be a dedicated circuit
> > (and probably a dedicated timeslot) on each and every one of those
> > switches)
> >
> > : or there is a dedicated line between one location & the internet
> > : backbone (ib) & from the ib to the other location?
> >
> > Not for a point-to-point line. But there are variations of that
> > approach such as ATM in which what one gets is dedicated virtual
> > circuits. Point-to-point T1's always connect the same two locations;
> > virtual circuits in some of the other technologies allow bandwidth
> > guarantees to be established for the duration of a session, with
> > the endpoints being determinable dynamically (provided the endpoints
> > are both in the service area of the technology.)
> >
> > : 5.How does one go about figuring out the cost of
> > :hardware,software,manpower & time to implement Voip in the company
> > :XYZ?
> >
> > One hires a consultant who has done it before. If you try to do
> > VOIP with people who are unfamiliar with the technology and haven't
> > had to do similar real-time work before, chances are excellent that
> > mistakes will be made, configurations will be experimented with,
> > the wrong equipment will be bought, the wrong dedicated line type will
> > be installed (on a multi-year contract), the internal switches won't
> > be upgraded to QoS, or won't be upgraded to handle enough simultaneous
> > channels... etc., etc.. So if you aren't hiring a consultant to
> > determine all these prices on your behalf, then whatever figure you
> > come up with yourself, you had better multiply by about 8 for hardware
> > and change the implimentation time to "person-years" where you had
> > "person-months" before.
> > --
> > Are we *there* yet??
--
Regards
Stephen Hope - return address needs fewer xxs