[AG-TECH] Scaled-down versions of AG

[Allan Spale]

Shahid,

EVL has created a scaled own mini AG node that is stable on a single
machine (that is, VIC an RAT don't crash the computer after a certain time
interval or after certain actions), though there are flaky things that
seem to make it less than perfect (please read: things are still being
tested and reconfigured for improved usability). 

We are using a single Windows 2000, mid-speed (700-900 MHz...can't
remember the exact spec) Pentium III with about 700+ MB of RAM.  We are
multicast enabled on the network that the computer is connected to.  We
have one video camera, one microphone, a self-contained echo cancellation
unit, some Soundblaster audio card (probably the 128-bit), and the
Hauppage (sorry about the spelling) WinTV card.

Installation of the AG software on the Windows machine went ok.  However,
we needed to obtain a stock-vic (not the DirectDraw version since the
Transmit option is disabled).  We have not been able to create a .vic.tcl
file on Windows to configure VIC to startup with certain options transmit
on startup at 800 kbps, 28 fps, etc.).  There are command line options
that allow minimal configuration.  Finding a version of RAT for Windows
was difficult.  We had previously downloaded some Windows version of RAT,
but the MICE site seems to only have source code for the most recent RAT
version, which means compiling the code into a Windows executable (which I
have not had the motivation to do yet).  Have not tried dppt yet. 

A major downside to the single machine implementation is the inability to
use the Virtual Venues web interface properly.  MUD room changes work and
that might be it.  RAT won't switch rooms through the web interface since
there is not the correct script available on the web interface to do this.
VIC does not currently work through the web interface, but that may be
because we are not using the DirectDraw, so some files may have to be
moved around.

So, basically, we are still optimizing things on this mini AG, but at
least it *seems* functional to serve as a backup if we have other people
using the room where the AG is.

It may be best to go with a two machine setup to make things function more
properly.  If you want a quick test and don't have multicast set up, check
out VRVS' web site at http://www.vrvs.org.  Their installation package
involves some dated versions of VIC and RAT and probably will only permit
one camera, one mic (although I have not explored this in great
detail...you may just have to register multiple times to get multiple
instances of VIC running, just a guess though).  A multicast bridge of
your choice is provided, but the network bandwidth requirements of the AG
are still needed.  There are links to AG Virtual Venues using VRVS' web
site, so you will have AG connectivity.

So, best of luck for finding the correct path to take.  If you find out
any information, EVL woul be interested to see how other sites have
created a functional mini AG.


Allan
EVL at UIC
node-op

On Fri, 11 Jan 2002, Shahid Bhatti wrote:

> 
> We at the CS dept of UCSB are trying to create multiple remote sites (on
> campus) for participating on Access Grid  meetings. These mini AG nodes will
> be able to receive all the incoming video streams and able to send at least
> one video stream. For achieving this goal we are visualizing the AG site
> with minimal hardware required (as compared to the fully operational
> standard AG node also present at UCSB). Looks like two machines with lots of
> computing power will can do the job. One machine will be running both VIC
> and RAT on Linux and the second machine will act as Win2000 Display machine.
> I just wanted to ping in this list to those guys who have successfully done
> it in the past and what are the odds involved. Your comments are welcomed
> and awaited, thanks.
> 
> -Shahid Bhatti
> CS Dept. UCSB
>