[Bioclusters] Re: SGE and checkpointing

[Hunter Matthews]

On Mon, 2002-05-20 at 23:11, Ron Chen wrote:
> --- Hunter Matthews <thm@duke.edu> wrote:
> > >> Blast on RH 7.2
> > >
> > >I've never tried this before, but this is what you
> > >can do --
> > >
> > >1. patch the Linux kernel
> > >This approach may be harder, but this will add
> > check-
> > >pointing support in the OS, that means you can
> > check-
> > >point all applications without recompiling.
> > 
> > I had looked around a month or so ago, and it
> > appeared that
> > checkpointing was in a pretty unmaintained state.
> > 
> > What do people recommend/use for checkpointing? 
> 
> The Linux kernel patch was submited to Linus serveral
> months ago, but it is not really useful to most
> users...
> 
> You can try one of the libraries available from
> http://www.checkpointing.org/
> 
> What kind of applications do you want to checkpoint?
> Any specific reasons to use the kernel checkpoint?

Some of our applications, like PAUP, we don't really have the source 
for. For the others, it simply looked like an easier solution to me.

Our applications tend to be the normal long running, single system, no
network ports open variety. Local files open, but we all our systems are
"the same" as far as the users are concerned - we use kickstart and NFS
to present identical images to users.

I should take another look at checkpointing.org, sounds like.


-- 
Hunter Matthews                          Unix / Network Administrator
Office: BioScience 145/244               Duke Univ. Biology Department
Key: F0F88438 / FFB5 34C0 B350 99A4 BB02  9779 A5DB 8B09 F0F8 8438
Never take candy from strangers. Especially on the internet.