From: Sean Dague on Friday, May 06, 2005 5:32 AM > > On Thu, May 05, 2005 at 02:12:25PM -0500, Brian Elliott Finley wrote: > > David, > > > > Let me say thanks for your well thought out email. Also, do you know > > Jan Lindheim and/or Mark Bartelt? > > > > Comments below... > > > > > > David Mathog wrote: > > > > >Greetings, > > > > > >I use systemimager and boel to replicate nodes on our beowulf. > > >In order to speed things up I first modified boel to include > > >"dolly", which is a daisy chain copier, > > One of the main complaints I've heard about chaining topology for servers > is > that the reliability now drops a lot. Any failure on any server cascades > through all the rest down the chain. > > Have you looked at the multicast stuff instead? That scales as O(1), and > doesn't suffer the multiple single points of failure.
In a previous life, I did some investigations on this very topic. The points are well taken on reliability and the impact of a node failure. The impact can vary, if it's a straight pipeline or a tree structure, but it's still there. Having said that, multicast can be quite problematic (i.e., impossible to get to work correctly) for some switches, and I'm not just referring to the consumer variety. So, for the industrial systems that prompted this study, we ended up keep both sets of tools available, hoping that at least one would work--our preference was to start with the multicast. -- David N. Lombard My comments represent my opinions, not those of Intel Corporation. ------------------------------------------------------- This SF.Net email is sponsored by: NEC IT Guy Games. Get your fingers limbered up and give it your best shot. 4 great events, 4 opportunities to win big! Highest score wins.NEC IT Guy Games. Play to win an NEC 61 plasma display. Visit http://www.necitguy.com/?r _______________________________________________ Sisuite-devel mailing list Sisuite-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sisuite-devel
