Hello Sergio, Most of the requirements described in C.26-C.29 exist on the TeraGrid. Many of our HPC resources provide node local, cluster wide, and even grid wide user, project, and/or job directories. Sometimes they are scratch directories. We also have situations where the path depends on the user/project/job and doesn't have the same value for everyone. We currently satisfy these requirements by defining standard environment variables that users and jobs can use to discover the paths to various directories. We have also discussed the need for a grid service for looking up these paths so that an external entity that is constructing a job can for example query for paths to use in the job they will submit. Hope this helps.. Regards, JP On Nov 17, 2008, at 9:52 AM, Sergio Andreozzi wrote:
Hi JP, Balazs et Al,
in the last part of the today call, we have discussed comments C.26- C.29: http://forge.ogf.org/sf/wiki/do/viewPage/projects.glue-wg/wiki/GLUE20Comment...
I would appreciate your comments/considerations as related to your working context.
In the spec, we should also better clarify the relationship between working area, cache, tmp/scratch/appDir in computing manager class. Also the connection to ComputingShare.MaxDiskSpace should be identified.
Cheers, Sergio
-- Sergio Andreozzi INFN-CNAF, Tel: +39 051 609 2860 Viale Berti Pichat, 6/2 Fax: +39 051 609 2746 40127 Bologna (Italy) Web: http://www.cnaf.infn.it/~andreozzi