It's not the kernel interface so much as the lack of client-side
caching, which is a consistency decision, not an implementation one.
Regards,
Rob
On Apr 16, 2010, at 9:48 AM, Emmanuel Florac wrote:
Le Fri, 16 Apr 2010 15:25:33 +0200
Christian Baun <[email protected]> écrivait:
Any ideas why it needs so much longer when using the Linux Kernel
Interface and how to benchmark the performance in a better way?
This is to be expected. The kernel interface is very suboptimal and
performs generally "just OK". "cp" particularly fares very poorly on
pvfs (that's why there's pvfs2-cp...). It works well enough to be used
in production, even with samba/nfs shares for instance. It works
really
well for FTP (and big files).
I fancy someday to see how the pvfs2 kernel module could be enhanced,
however I'm missing both the ability and time ATM :)
--
------------------------------------------------------------------------
Emmanuel Florac | Direction technique
| Intellique
| <[email protected]>
| +33 1 78 94 84 02
------------------------------------------------------------------------
_______________________________________________
Pvfs2-users mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users
_______________________________________________
Pvfs2-users mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users
