All things related to multicast

["Lawrence A. Rowe" <>]

Hi -

We are producing a webcast of the U.C. Berkeley Systems Seminar this
week - thursday Nov 29, 2001 4:00-5:00 PM - at which Prof. Dave
Patterson will give his widely acclaimed advice on being a faculty
member lecture.  A title and abstract for the talk and instructions on
how to watch the webcast are given at the following web page:

        http://bmrc.berkeley.edu/courseware/misc/01-11-29/

Note the link to a web-based text-only chat room that will be used for
reverse questions and production problems.

This will be the first webcast we have produced this semester, I'm on
sabbatical this year so the Berkeley MIG Seminar is not being offered. 
We have been making a lot of changes in our webcasting infrastructure
that we will be using for the first time in this webcast.  Details are
given below.

The primary webcast, which will be composed of three video streams, will
use the Internet Mbone tools. We might simulcast using Real Networks
technology, but that is still to be worked out. Details on connecting to
the webcast(s) are given on the web page above.

You will need to use the new Open Mash binary tools (vat and vic) to
receive the webcast - we fixed bugs in existing vic related to decoding
large mjpeg images which are used in the new content stream.  This code
can be downloaded from http://www.openmash.org/users/downloads.html.

If you have problems during the webcast, contact
.
        Larry
-- 
Professor Lawrence A. Rowe          Internet:  

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

-----------------------------------------------------
Changes in the Berkeley Webcasting Infrastructure
L.A. Rowe (11/27/01)

We have been changing the physical infrastructure in Soda Hall to
support more classrooms and to improve the quality of the webcast
productions and reduce the operational overhead.  Specifically, we now
send four analog a/v signals from each classroom back to a broadcast
center where various a/v and computer equipment is located. In addition,
we have redone the audio support in 405 Soda to fix problems with noise
from the projector and threshholding of questions from the audience. 
You should hear audience questions from all sides of the room and the
audio should be cleaner.

The new webcast infrastructure includes an RGB capture board that can
digitize the RGB signal directly from the presentation computer, as
opposed to scan converting the signal to NTSC and then digitizing it. 
The RGB stream is sent using MJPEG at 1 frame per sec - the image size
is 1024x768.  The images are considerably better than the scan converted
NTSC streams, but they do take a lot of bandwidth (approx. 2 Mbs) and
require a lot of processing to decode.

The new infrastructure allows us to send more streams.  We will send 4
streams for the Patterson seminar: RGB content, speaker/audience, stage
view, and NTSC content.  You can mute the RGB or NTSC stream so you only
need to watch one.

We have also installed a Parkervision tracking camera for the close-up
view of the speaker.  We will try it for the webcast, although the image
quality is not as good as the less expensive pan/tilt cameras (e.g.,
Cannon vcc3, etc.).

The most important change is the installation of the new Cisco-based
network which should dramatically improve multicast performance.  We've
been doing experiments locally and the new network runs much better than
the Nortel equipment used in the rest of EECS.  We have noticed one
problem with multicast -- every 15 minutes or so, the multicast packet
transmission freezes for about 60 seconds, after which it resumes.  This
is probably something to do with the way we have the equipment installed
and the Cisco code we are running.  We need more time and help debugging
this problem.

Finally, we will experiment with a new transcoding gateway that produces
Real Networks (RN) streams.  It combines two streams in image space
before passing it to the RN encoder.  We will simulcast this session to
test this new code.