All things related to multicast

[Hugh LaMaster <>]

On Tue, 4 May 1999, Danny McPherson wrote:

> Cool!  Any idea how many receivers and the max number of time a single 
> router
> had to duplicate a "stream"?  Thanks!

I didn't explain the application much last night, since I'm pretty
much focused on the router aspects, but, the max number of receivers 
we have used is fairly small, because this is actually a combined 
unicast/multicast application with low-bandwidth application control
information potentially flowing from the clients.  The project didn't 
build a receive-only version, since it really is designed for multiway 
collaboration. Given that we had only live, active clients, we had to
limit participation/distribution of the clients, since we couldn't
really have random participants manipulating the images.  Apologies
to those who requested participation which we couldn't accommodate
for that reason.  All the non-internal sites had active participants.

So, because we limited distribution, there were four sites outside 
of Ames participating: UC Santa Cruz/(also hosting Salinas Valley
Medical Center), Stanford Univ. Med. Ctr/Biocomp center, NASA 
Glenn Research Ctr., and the Navajo Med. Ctr. in Shiprock, N.M.,
as well as internal receivers at ARC.  (We were unable to get a
full-bandwidth link to to Navajo Med. Ctr. in time, due to 
logistical and weather problems, although that site was able
to participate in the low-bandwidth portions of the test.)  The
focus for this demonstration was to use a real-life collaboration
application with a limited number of participants.

The max number of times any router had to duplicate a multicast 
stream was three: one input stream ingress via full-duplex ethernet, 
three egress streams via ATM subinterfaces on an ATM.  This was on
the router which attached to the source LAN.  The minimal replication 
was by design, since most of the configuration was in place to support
50 Mbps, which we used during testing.  (Basic arithmetic: with a 
single ATM interface, the total replicated traffic is limited by
the bandwidth of the port.  All the NREN links and ports are OC-3,
although there were some OC-12 links downstream.)  The host attachment
was via full-duplex fast ethernet.  On the WAN side, most of
the routers forwarded from one input to one output, with a couple
routers duplicating streams.

-Hugh



--
 Hugh LaMaster, M/S 233-21,    ASCII Email: 

 NASA Ames Research Center     Or:          

 Moffett Field, CA 94035-1000  No Junkmail: USC 18 section 2701
 Phone: 650/604-1056           Disclaimer:  Unofficial, personal *opinion*.