Cell Relay Archive

[albert.e.manfredi@boeing.com]

In article <3749B6DE.4EBCCDF6@gte.net>,
  Baris Aksoy <baris.aksoy@gte.net> wrote:
> Hello,
>
> As I understand, there's debate going on about VoIP vs. VTOA.
Actually,
> I have two points that I'm not sure about:
>
> 1) When people mention Voice over IP, do they mean to put the voice
> directly in IP packets and send it directly on fiber/copper with PPP
> protocol?

Yes, essentially, except that provisions are made to give the voice
packets higher priority in the queues at each hop. This even could
include higher priority at the link layer, within a LAN, not just
between routers. Although that link layer stuff is independent of IP.
But it is part of the whole picture.

>                     OR
> do they mean to put the voice in IP packets and then put these packets
> in ATM
> cells, to be able to send it on ATM backbone? If they do the latter
way,
>
> can ATM handle QoS for voice effectively, even the voice is in IP
> packets ?

Well, that too, if you are doing IP over ATM. And this might even
become common, although it really seems twisted to me. Doing IP over
ATM pretty much spoils any ATM QoS advantages you might have had. The
ATM Forum is just now getting into ways of intelligently applying IP
differentiated services over an ATM backbone. Still, my feeling now is
that any IP over ATM can only degrade the QoS advantages ATM promises.
But overall it could still become the best solution.

> 2) IP packets are not effective as ATM cells to carry voice which is
> delay sensitive. Do you know any kind of researches/plans going on to
> use different IP packet size for real-time data?

This has been done already. An excellent article is in the Proceedings
of the IEEE, December 1997, but it's at work. There are likely others
that are more recent. The optimum IP packet size, over a packet-
switched medium, for voice over IP varies according to a whole lot of
variables, such as distance (no. of hops), speed of backbone, loading
of backbone, and so on. A typical number is something on the order of
200 bytes per voice packet.

--
Bert
manfredi@arl.bna.boeing.com


--== Sent via Deja.com http://www.deja.com/ ==--
---Share what you know. Learn what you don't.---