[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: high-quality mjpeg video streams using ip/rtp over internet2

Hi Denis -

Thanks for the feedback.  As you know, Matt has been the one primarily
doing the work -- which is much harder than most people realize.  We
have been trying to make this application run as well as possible on the
Internet.  That means no excuses for lip sync, dropped packets, strange
hardware (e.g., audio chips that don't really give you 16K samples per
second like they say they do, and so forth).  My guess is that you could
put together a send/receive box for something around $1.5K no including
a monitor and all the other goodies you get on a desktop. (We call this
a "delcobox" after Matt :)!)  Eventually, it ought to cost under $500
but that's still in the future.  We'd be happy to work with you to setup
such systems if you want to do it.

On the 1 second latency issue, that's a constant that Matt has set in
the software.  I think we can do reduce it to something under 1/2
second, but we still have to experiment with that and see what effect it
has on packet loss and so forth.  Personally, I am anxious to find out
if we can run it under 300 msecs which is the upper bound for normal
interactive talking.

I should also tell you what other things we are working on here at
Berkeley.  We are doing experiments to see if we can build a
software-only player that can receive and play the rtptv (we renamed it)
streams.  We think that we can use current hardware to do 3-5 Mbs CIF
images where the software drops the 2nd video field and line doubles the
1st field. We are putting this into Open Mash, so it will be available
for many other Mbone and streaming media applications.  We still have
quite a bit of work to do to see if that statement is true. If so, we
might be able to combine a delcobox with hardware and software playback
at different places.

Finally, I'm pushing hard to get our new network switches installed,
which is the limiting factor to us producing the MIG Seminar webcast we
both want to see.

Thanks again for the feedback and I join you in applauding the job Matt
has done developing this software.
	Larry 
-- 
Professor Lawrence A. Rowe          Internet:  Rowe@BMRC.Berkeley.EDU
Computer Science Division - EECS       Phone: 510-642-5117
University of California, Berkeley       Fax: 510-642-5615
Berkeley, CA 94720-1776            URL: http://bmrc.berkeley.edu/~larry


Larry and Matthew,

I did setup a little demo of rtpvb for Chris Thomas and Ed Sakabu,
engineers with the Academic Technologies Services unit here, this
afternoon... everything went well, and I dare say they were properly
impressed. They have some questions and comments regarding applicability
of rtpvb for some video-bridging scenarios here at UCLA... I hope thru
this email to introduce you both.

For me two new things. This was my first time to see rtpvb's output
displayed on a real TV... it looked great. Nearby, Ed fired up another
monitor and tuned in to CNN on UCLA's cable TV system... of course we
played the game of flipping between rtpvb and the local CNN signal and
barring the 1-second delay to get/process the Berkeley video feed we
couldn't -- of course -- tell them apart.

The other thing that tickled my fancy was that their lab network is
connected very close to the UCLA Calren2 GSR and so the path to Berkeley
was extremely clean -- the video client for example was reporting
essentially "zero" loss, 0 to 2 pkts per 5-second periods. This was
between 3 and 4pm this afternoon... we were pulling an aggregate of close
to 11-Mbit/s. This is the cleanest path to UCB and BMRC that I have seen
ever. So we do have excellent networking infrastructure and capacity to
support apps like rtpvb... hopefully, someday we'll have multicasting
working here as well. Ed tickled the ethernet connection to cause brief
net outages but the rtpvb clients handled the gaps gracefully indeed... so
all in all, terrific!

Thanks again Matthew for your help and congratulations for the continued
work to make rtpvb such an exciting application. In fact I have become so
spoiled with rtpvb that seeing the Multimedia Seminar last Wednesday in
220-Kbit/s RealVideo was a real downer... so I look forward to when the
seminar can be broadcast experimentally with rtpvb.

-- Denis