[gin-jobs] Standarized Execution Environment
Thomas Soddemann
soddemann at rzg.mpg.de
Wed Mar 28 10:56:27 CDT 2007
Hi Laurence,
I had a quick look at the proposed variables. They look reasonable to
me, although I see a technical problem in supporting GRID_USERID and
GRID_GLOBAL_JOBID in an infrastructure like DEISA.
In principle I favor a separation of concerns, meaning that a local
resource management system does not need to know about the origin of a
job. This is taken care of by components like GRAM or the UNICORE NJS.
But maybe I am overlooking something.
Why do you guys think that a job should be aware of those environment
variables?
Why should a job know at this stage that it was started as a grid job in
first place? (Ok, a grid-enabled job should avoid exploiting local
peculiarities, and how to do that is e.g. in DEISA formulated in a
policy document.)
Thanks,
Thomas
Laurence Field wrote:
> Hi,
>
> One of the problems that we found during the interoperation activities
> was with the different environments found on the worker nodes in the
> different grids. Getting the jobs to arrive on the worker node is one
> problem, ensuring that the jobs run correctly is another and this
> problem has mainly been overlooked. During the joint operations
> workshop in Bologna between OSG and EGEE this problem was brought up and
> discussed. The result was the following document which suggests
> standardizing the execution environment.
>
> http://edms.cern.ch/document/630962
>
> I have already scheduled a BOF at OGF 20 to discuss this issue further.
>
> Feedback on this topic is more than welcome.
>
>
> Laurence
> --
> gin-jobs mailing list
> gin-jobs at ogf.org
> http://www.ogf.org/mailman/listinfo/gin-jobs
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: soddemann.vcf
Type: text/x-vcard
Size: 345 bytes
Desc: not available
Url : http://www.ogf.org/pipermail/gin-jobs/attachments/20070328/8b4fee91/attachment.vcf
More information about the gin-jobs
mailing list