Filesystem

Angel Rivera angel at wolf.com
Sat Aug 2 08:48:09 EDT 2003


Steven Timm writes: 

> I have some anecdotal evidence that ext3 starts taking performance hit
> in cases where there is a lot of files getting written and then
> quickly erased.  Also there's a performance penalty on burst I/O--e.g.
> if you have a system doing near-continuous disk writes and reads it
> will bump the load factor up.  But I don't have any information to
> suggest that Reiser does it better. 
> 

It depends what you are going to use the nodes for. For normal compute 
nodes, I don't think there is enough of a payback to change ext3. For our 
disk nodes, we use ext3 for system filesystems and XFS for the exported disk 
space (with NFS patches and tuning of couse) to get some serious 
performance.  We are currently testing different filesystems on one of the 
disk nodes we just purchased and have seen a dramatic rise in performance 
with the above. 
_______________________________________________
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf



More information about the Beowulf mailing list