[51576] in North American Network Operators' Group
Re: Broadening the IPv6 discussion
daemon@ATHENA.MIT.EDU (Petri Helenius)
Fri Aug 30 04:28:47 2002
Date: Fri, 30 Aug 2002 11:28:09 +0300
From: Petri Helenius <pete@he.iki.fi>
To: itojun@iijlab.net
Cc: nanog@nanog.org
Errors-To: owner-nanog-outgoing@merit.edu
itojun@iijlab.net wrote:
>
> you can go hybrid, like
> - client connects to server for game playing info (like location on the
> map, inventory and stuff)
> - client will talk with each other directly for video/voice-chat
> even with this, server load/traffic will be decreased.
This is exactly what I also had in mind. This would get 1:10 benefit
in bandwidth and actually enable this kind of activity.
>
> i still don't understand why you say multicast is mandatory.
>
Most consumer connetions (where this is feasible anyway) are asymmetric,
having 256k-1.5Mbps downstream and 128k-512k upstream. A decent video stream
represents 128k to 384k of bandwidth. If you have a small number, say eight
players in a game, you'll end up sending the stream seven times unless
you do multicast. You probably don't have the upstream bandwidth to accommodate
that unless you're lucky to sit on top of a new housing development with
fiber in the basement.
The next logical step to this discussion is what happens to multicast routing
when one million gamers setup half a million *,G and a few million S,G pairs.
Add a zero if it makes the excersise more interesting. Keep in mind that
one million gamers playing is less than what the network currently has at any given
moment.
Pete
