Multicast status

Tony Rimovsky tony at ncsa.uiuc.edu
Wed Feb 9 10:20:34 CST 2000


Tom, please don't pursue your own bridge.  I am fairly certain that you will
run into insurmountable technical difficulty getting one to work, and I
can't devote the resources to helping you do troubleshoot.  Setting up a
bridge at access-dc will increase my group's workload and be in direct
contradiction to what the access-grid developers are trying to accomplish.
There is also a good possibility that the current pipe to access-dc couldn't
handle the load..

Currently, the bridge is not simply a backup.  It discourages some of the
core goals of the access-grid model by allowing remote sites to be lazy
about dealing with multicast.  After a year of using the access-grid, there
are still active sites no closer to native multicast because, in part
because the bridge exists.  Furthermore, the bridge confuses
troubleshooting, further complicating the multicast picture when something
does go wrong.

There is no reason for this community to allow something as significant as a
Chautauqua happen again without having put the advance effort in to make
sure that key sites are enabled.  Smaller impromptu events are taking a risk
in relying on access-grid technology at this stage.  All it takes is one key
person to be unavailable for things to break down (and that is true even if
you don't use multicast.)

Bill has outlined alternative solutions which encourage the proper
direction.  He has offered to help sites. I extend the same offer.
Particularly with the NSF, it is significant that we we get proper multicast
working, either natively or using specific hardware at the remote sites.

And, in point of fact, not every major event has required using the bridge.
The recent Illinois Governor's visit is a positive example, and I believe I
heard Larry describe it as the best access-grid demo yet.  As time passes,
the technology has improved.  Troubleshooting multicast problems is almost
routine now, with the bigest delays being caused by unavailable staff, not
mysterious technology failures.

Final point: if the community decides that a backup mechanism is necessary
as a backup, then ANL should organize the implimentation, _not_ a rouge site
that decides they know better.  Make your case to ANL people and the rest of
the community.  If people generally think you are making sense about
backups, then I am sure a bridge of some sort will remain.  (They may even
decided it should be hosted at ACCESS!)  An open invitation was part of the
propsal sent to the list on 1/27 and there hadn't been any negative feedback
(other than some concern about increased $$) in response until now.  The
point about backups probably needs to be discussed.  Bob/Bill/Lisa, do you
have anything in mind with respect to a safety net?



On Wed, 09 Feb 2000, Tom Coffin wrote:

>  
> True, the mission of the Access Grid is to exist in a multicast 
> environment.  
> However, for large scale production events like a Chautauqua 'bridges'  
> should still be in place and accessible as 'backup'. (for example every 
> large event [sc99, chautauquas] to date has required ACCESS to switch  
> to 
> a bridge to be seen or heard. We employ telephone backups why not 
> network backups? 
> 
> regarding cise-nsf.gov, they currently do not have multicast enabled. 
> I believe they want a permanent presence on the Access Grid project.  
> I will touch base with the folks at NSF to see what's going on with 
> their becoming multicast enabled. 
> 
> Where can I learn more information on setting up a multicast bridge? 
> 
> ____________________________________________ 
> At 09:21 AM 2/9/00 -0600, Robert Olson wrote:  
> >>>> 
> 
> > In the effort to rid ourselves of multicast bridges, I thought it'd 
> > be good to get an update on status from folks. The following groups 
> > are in the bridge configs: 
> > 
> > VA Linux        Corporate folks, no working multicast 
> > Utah            I seem to remember you guys having working 
> > multicast 
> > Boston          Did you get your one-way multicast fixed? 
> > UNM             This is a work in progress, right? 
> > MPHCC           Also a work in progress if I recall correctly 
> > Kentucky        Also a work in progress if I recall correctly 
> > cise-nsf.gov    Tom -- this was a one-time thing, right? 
> > EVL/UIC As I recall, this requires a longer-term fix. Status? 
> > UIUC CS dept    onetime for Dan Reed I think. Anyone know status  
> >                 of multicast there? 
> > Nestor's lab    Bill, I think this is working multicast now, right? 
> > microsocopy.com Nestor's home. hm. 
> > msu.ru  Onetime demo, I think 
> > 
> > thanks, 
> > --bob  
> > 
> <<<< 
> 
> 
> 
> 
> 
> 
> ___________________________________________________________ 
> Tom Coffin .......................... tcoffin at ncsa.uiuc.edu 
> 

-- 
 +-------------------------------------------------------------------------+
 | Tony Rimovsky, Manager -- Network Development    Phone: +1 217 244-4728 |
 | National Center for Supercomputing Applications  FAX:   +1 217 244-1987 |
 | 605 E Springfield, Champaign IL, 61820           tony at ncsa.uiuc.edu     |
 +-------------------------------------------------------------------------+



More information about the ag-tech mailing list