It's a good idea to open up a new browser window to be able to look at the map at the same time as you read.
And this time it has to work. But it doesn't, and it won't with the current configuration.
Now why is that so?
Simply because no multicast traffic is allowed to pass between magda-gw
and inner-gw to isolate the MERCI MBone testbed (see picture below)
, which is also running through magda-gw, from the public MBone running
in the rest of the world.
Here's a picture of the
MERCI MBone
testbed.
If we are going to run public MBone to Karskrona/Ronneby and Mitthögskolan, and run multicast over the link we have to be sure that the subnets on the Stanford side are completely isolated from the Stanford University network (SUNET). This to avoid routing irrelevant traffic over the link.
I can see four more or less possible solutions:
The best solution seems to be to get a separate router for
S-SVL, hook it up on the FDDI ring on one end and the Stanford link
on the other and let the MBone traffic flow.
Thereby skip the cludge 2x10BaseT that we no longer need
now that we only have one T1 speed anyway, and above all take some load
off magda-gw.
Here's a picture of the
proposed solution and a picture of the resulting
multicast routing network.
We have a router that have a FDDI interface soon hooked up on the
Electrum-ring and can handle multicast routing. It is called eit-gw.
Unfortunately it has no slots left for the link, and is also routing
some other things(sys-adm course?) so we still need a separate
s-svl router.
The second best solution is to isolate the MERCI MBone even more by hooking it up on a router of it's own, and open up magda-gw for public multicast traffic again. This means we will still have the famous cludge 2x10BaseT in the system as well as more parts in the system that can break
A third solution is to move the Stanford link to inner-gw, if possible. This will not give any protection against random load fluctuation generated by other subnets attached to inner-gw. This was not possible due to lack of slots in inner-gw.
The fourth solution is to do the UCL variant and put one "switch" between inner-gw and magda-gw and another between magda-gw and sanfran. Thereby we can swith the network between MERCI-mode and S-SVL-mode. Panos would hate me for suggesting such a thing ofcourse, because it would shut down a third of the MERCI testbed everytime someone wants to talk over the stanford link.
Maintained by Tobias Öbrink