[streaming] Re: Venus 2004 : which architecture ?
Francois Meyer
fmeyer at obs-besancon.fr
Fri Jan 23 11:26:11 CET 2004
On Thu, 22 Jan 2004, Benjamin PRACHT wrote:
> Hello, here are 1 or 2 ideas about your project :
>
> - The archtecture you are panning makes sense to
> me. However, be aware that you'll need a quite
> recent computer tu plug the webcam in : you'll
> need to make real time encoding, which requires
> some cpu power (p4 1.5 GHz for real time
> fullscreen MPEG4 encoding).
Is there any alternative encoding option that would
realize a compromise between quality and CPU needs ?
In fact the transmitted image will be rather steady
and will have a good response to compression (the
reason why I'd like to stream it live is because of
the air turbulence that will really render like a
real telescope view ; though, steady, the image will
be alive).
The webcam full res capability is limited to 640x480
but I would like to stick at least at 320x240 if
possible.
> - which way will you use to stream on the internet
> on the servers ?
Well, that is part of my question :) I am open to
any suggestion on that topic. I am really new to
streaming and I barely have a small idea of what is
possible to setup in such a case.
> Using vlc as a udp multicast -> http proxy would
> make sense, in my opinion. However, be aware that,
> as far as I know, no high load test, with lots of
> http connections have been made on vlc. In fact,
> we would be pleased to get some feedback on that
> point
I will be trying to make some test runs as soon as I
have the basic bricks to build the system. June is
still far, but a lot of work has to be done to
qualify the stuff. Your help will be REALLY valuable.
>
> - vlc is unable to command a dv camera directly.
> You will need to use an external program to set
> the camera, then have vlc read the resulting
> video directly on a device / fifo, and transcode
> the result in a streamable format. This should
> be possible, but hasn't really been tested,
> AFAIK.
Ok. The DV item is a 'cherry on the cake' option, in
case we quickly solve the main problem. Anyway
controlling the camcorder is not really an issue :
the camcorder is not intended to be remotely
operated ; grabbing the stream is the only thing I
intend to do with vlc.
I summarize the key points of the problem at this
stage and my degree of understanding of how to
address them :
1. grab the webcam stream ;
Ok, no problem (I think).
2. encode the stream :
need your help on the format to be used
keeping in mind the constraints I developped
in the beginning of my message (320x240, 5
im/s, rather steadyi, high contrast image,
and no audio needed at least in a first
stage...)
3. multicast the result to the server(s) :
I think I could go through this one, though
your help is still welcome.
4. provide access to the stream from the internet :
select some clients that we want to be able
to read the stuff, or stick to vlc ?
according to he above result, select the
method to provide access to the stream.
5. evaluate the per user bandwidth, CPU
requirements, number of servers needed ;
this will be done once point 4 is solved,
getting feedback from the first test that
I'll try to setup asap.
Ok that is enough for now.
I will start to address point 1 and 3 adopting a
generic solution for point 2 just to setup the first
tests.
I am carefully listening to your thoughts about
all that.
Thank you again and in advance.
-- Francois Meyer (Please, no html, no MSWord attachment)
Tel : (+33) 3 81 66 69 27 Fax : 3 81 66 69 44
Observatoire de Besancon - BP1615 - 25010 Besancon cedex - FRANCE
**** Université de Franche-Comté ****** CNRS UMR 6091 *****
--
This is the streaming mailing-list, see http://www.videolan.org/streaming/
To unsubscribe, please read http://www.videolan.org/support/lists.html
If you are in trouble, please contact <postmaster at videolan.org>
More information about the streaming
mailing list