You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is there some way to greatly reduce the number of files of the pattern $A/orthofuse/$B/.groups and $A/orthofuse/$B/.groups.orthout generated by oyster.mk?
I'm a support analyst at an HPC installation, and I have a user with an ORP job that has generated over half a million files in that one directory. The typical size of a file is just a few hundred bytes. This is absolutely brutal on our shared network filesystem, it has negative performance consequences not just for the user running ORP but for all the users of the filesystem.
Best practice in the HPC world is to expose the filesystem to a few large files rather than a lot of small ones. Is there some way that ORP can be invoked that will do this? If not, can we make it a feature request to find some way to minimize that file-count? Using a database (like SQLite) instead of the filesystem, say, or using incremental tar or dar to stash the *.group and *.group.orthout files?
Thanks,
Ross Dickson PhD, Support Analyst
Compute Canada / ACENET / Dalhousie University
The text was updated successfully, but these errors were encountered:
Is there some way to greatly reduce the number of files of the pattern $A/orthofuse/$B/.groups and $A/orthofuse/$B/.groups.orthout generated by oyster.mk?
I'm a support analyst at an HPC installation, and I have a user with an ORP job that has generated over half a million files in that one directory. The typical size of a file is just a few hundred bytes. This is absolutely brutal on our shared network filesystem, it has negative performance consequences not just for the user running ORP but for all the users of the filesystem.
Best practice in the HPC world is to expose the filesystem to a few large files rather than a lot of small ones. Is there some way that ORP can be invoked that will do this? If not, can we make it a feature request to find some way to minimize that file-count? Using a database (like SQLite) instead of the filesystem, say, or using incremental tar or dar to stash the *.group and *.group.orthout files?
Thanks,
Ross Dickson PhD, Support Analyst
Compute Canada / ACENET / Dalhousie University
The text was updated successfully, but these errors were encountered: