Archives


- Beowulf
- Beowulf Announce
- Scyld-users
- Beowulf on Debian

batch systems with job deps (afterok)

Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.

Search

Jakob Østergaard jakob at unthought.net
Sun Sep 30 01:57:04 PDT 2001


On Wed, Sep 26, 2001 at 11:03:10AM +0200, hanzl at noel.feld.cvut.cz wrote:
...
> Typically, previous invocation of 'make' would still run when I have
> the next steps ready. Typing "make" again at this moment would cause
> both copies to work on unfinished step (second make would not wait as
> it should). (Typically, I might finish the next step on friday and
> would like it to be invoked during the weekend, as soon as possible.)
> However there might be a simple solution (like one 'make' in loop,
> until there is nothing to do).
> 
> 
> I wonder why my needs seem to be uncommon - is it because you guys on
> the beowulf list
>  1) are not as lazy as I am and you have the whole program ready in time?

If you believe that one, I have this big tower in london that I can sell
you really cheap - it's got a clock on it too  ;)

>  2) you change data/parameters rather then programs?

I believe it is rare to change your program, and have the new "incarnation"
depend on results from an older incarnation (eg. obsolete or at least different
version).

In fact, I can't imagine how you manage to keep track of how you got which
results from which code, with a pipeline of various incremental versions of
your software depending on results from other versions.   But maybe that's just
me   :)

>  3) you have plenty of time and do not need to work in parallel with
>     your cluster?

Well, most scientists don't have a lot of work to do, and besides, you 
need time to spend all that money too...   Eh...    ;)

> 
> I do not beleive any of these, please tell me why you all are not
> calling for job dependencies in any job spooling system :)

I think (2) comes fairly close to at least some of the truth.

Of course you change your algorithms.  But I have never worked with
multiple revisions of code depending on results from other revisions.
And I have never heard of anyone else (except you) doing so to such
an extent.

-- 
................................................................
:   jakob at unthought.net   : And I see the elder races,         :
:.........................: putrid forms of man                :
:   Jakob Østergaard      : See him rise and claim the earth,  :
:        OZ9ABN           : his downfall is at hand.           :
:.........................:............{Konkhra}...............:




More information about the Beowulf mailing list