Le 2015-11-17 11:15, Martin Bähr a écrit :
Excerpts from Maxime Boissonneault's message of 2015-11-17 14:29:28 +0100:
On a cluster, having tens of thousands of files will be a killer, but
even on a personnal computer, it will make your data management a
nightmare.
I have seen users who think that having 2.5M files in a directory is "ok"...
i'd like to side with the users on this.
it should be ok, and if filesystems were designed like databases then 2.5M
files would be ok, just like having a database table with 2.5M records is ok.
it is not the users fault that filesystems were not designed for their needs,
and while the users should be taught how to cope with this, i consider their
expectation that having 2.5M files in a directory is ok, in itself is actually
reasonable.
While I sympathetize with users on this, the reality is still there, and
will bite you in the ass if you don't take it into consideration.
Filesystems are simply not databases. They don't have configurable
indexes, they cannot be sorted, etc.
A filesystem is simply the wrong tool to use if that is the problem they
are having. We should teach them to use
the proper tool for a given problem.
The same is true with databases by the way. Sure, you can easily have
tables with millions of entries. But if you
use the tool wrong and don't configure indexes for example, your
requests will under perform.
Cheers,
Maxime
_______________________________________________
Discuss mailing list
[email protected]
http://lists.software-carpentry.org/mailman/listinfo/discuss_lists.software-carpentry.org
