Esteban,
All the I/O servers are striped across by default. If you want to
limit the number of servers striped across, you can do so by setting
an extended attributed on the _directory_ where you create the file:
setfattr -n user.pvfs2.num_dfiles -v "4" /pvfs/my-dir/
Alternatively, you can set the default stripe width for all files in
the filesystem with a config option in the file server config:
DefaultNumDFiles 4
This goes in the <FileSystem> context.
-sam
On Aug 5, 2008, at 6:18 PM, Esteban Molina-Estolano wrote:
Thanks for your help!
I think I was unclear about the number of I/O nodes, though. I
looked at
the filesystem config file generated by pvfs2-genconfig and I can
see the
list of I/O nodes, but I'm not looking for the total number of I/O
nodes.
Rather, I'm looking for the per-file metadata setting that
determines how
many I/O nodes the file is actually striped across. Where can I find
how
this value is set by default?
Thanks,
~ Esteban
stripe_size is in include/pvfs2-dist-single-stripe.h
the base IO node is determined in a round-robin fashion so 10 files
will ideally start on 10 different servers, if you had 10 servers.
number of io nodes comes down through the distribution, and initally
from the filesystem config file, I believe.
Kyle
On Mon, Aug 4, 2008 at 2:55 PM, Esteban Molina-Estolano
<[EMAIL PROTECTED]> wrote:
I'm researching placement algorithms for different large-scale
filesystems
using trace-driven simulation, and I'm trying to write a simulator
for
the
PVFS simple stripe placement algorithm.
There's one detail I haven't been able to find in any
documentation: how
the values for the base I/O node number, number of I/O nodes, and
stripe
size are selected, by default, for each file.
So far I've looked through the docs on the website, the dev mailing
list,
the 2000 paper ("PVFS: A Parallel Virtual File System for Linux
Clusters"), the document "PVFS2 Distribution Design Notes", and the
simple
stripe placement code (dist-simple-stripe.c).
Where else should I be looking for these default settings?
Thanks,
~ Esteban Molina-Estolano
_______________________________________________
Pvfs2-developers mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers
--
Kyle Schochenmaier
_______________________________________________
Pvfs2-developers mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers
_______________________________________________
Pvfs2-developers mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers
