Re: OpenMosix and Checkpointing

Daniel-

> It seems that, to migrate a process, you must transfer its state
> (address space, register file, etc.) from one machine to another.  If
> you can do that, it seems you could 'intercept' the transfer, and
> save/restore that state to/from a file.  Of course, my understanding
> of openMosix might be totally wrong (again, I've only just started to
> look at it), so please correct me if I am, ad I want to understand it.
>
OpenMosix does quite transfer the programs in total. It still has part
of teh process left on the orginal host. I am a bit rusty on the
details so will not tell you something totally wrong, but the webstie
does have some good papers on exactly what is happening.

>
> The work I'm discussing is not HAC (e.g. automatic, transparent,
> fail-over of a webserver, etc.) but rather providing fault-tolerance
> for HPC applications, i.e., if an application fails 15 hours into a 30
> hour run, we can (maybe after the whole cluster is rebooted) restart
> the application from a saved checkpoint file and not have to re-do all
> the lost work.
 This is very useful, good luck with it. the CHPOX is a good place to start.


-------------------------------------------------------
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&kid0944&bid$1720&dat1642