Re: OpenMosix and Checkpointing

Daniel Marques wrote:


Hi Dan,

> I'm a computer science researcher who has been studying checkpointing.

Last year I worked on a fourth year undergraduate group project looking 
into taking consistent checkpoints on an openMosix cluster using CHPOX, 
so hopefully I can answer a few of your questions. :-)

> It seems that if OpenMosix supports process migration, it should be
> possible to extend it to support checkpointing.  Is that what CHPOX
> does, or is it an independent approach to taking checkpoints?  

CHPOX is an independent approach to taking checkpoints, that will run on 
an openMosix cluster, as well as a standalone machine.  I don't think 
there is much (if any) difference, as far as CHPOX is concerned, between 
taking a checkpoint on a standalone, and taking a checkpoint on an 
openMosix cluster.  From what I remember though (and it was around a 
year ago since I did any of this), a migrated process is migrated back 
to the home-node before it is checkpointed.

For what it's worth, our solution involved pausing all communicating 
processes, checkpointing them, saving the contents of any message queues 
the processes were using at that time, and then restarting each of the 
processes.

Not particularly efficient - there are much more efficient algorithms 
out there!  But it was good enough for an undergraduate project!

One gotcha, if you use CHPOX - we found it to be quite unstable when we 
used it.  Having said that, a year has passed and it may be much better 
now! :o)

HTH,
	Tom



-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642