NAS + NFS

Jerker Nyberg jerker at update.uu.se
Tue Apr 30 05:20:01 CDT 2002


Hello,

Thanks for your answer. Do you know what the performance is for XFS when
greating many files? There was a discussion in LysKOM conference system
http://www.lysator.liu.se/lyskom regarding a comparison between different
filesystems and creation of files in them.


If I remember correctly Solaris/VxFS was the best closely followed by
Linux/ReiserFS (but a too new filesystem to be know as stable yet),
OnTap/WAFL (Netapp) had a quite strange limit (maybe maximum 10MByte
directory), Windows/NTFS (Windows) is ok. Some other Unix-filsystems
(UFS/Ext2/Ext3 etc) are quite bad and get too slow after creation of a
couple of hundered thousand files.  OS/2 with HPFS was quite ok even on
old hardware.

Here below is a small program for creating a million files in the current
directory. Could you or somebody else that have access to a Linux machine
with XFS try it out to see how long time it takes? As a comparison, on
ReiserFS it takes under ten minutes and on Ext3 you don't want to wait for
it to finish. I tried it on GFS, it took an hour or so if I remember
correctly.

Regards,
Jerker Nyberg.

- - -
#!/usr/bin/perl
$files = 1e6;
$a = "a";
push @name , $a++ for 1 .. $files; # Build array of names
print scalar @name, " names generated.\n";
print "Starting to create files at ".(scalar localtime)."\n";
$i=0;                           # Counter for progress bar
$|=1;                           # Un-buffer STDOUT the old way
for (@name) {
    print "*" if ++$i%(($files/80)+0.5)==0; # Print progress
    open F,">$_" or die "\nCreate failed: $!\n";
}
END{print "\nFile creation finished at ".(scalar localtime)."\n";}
- - -



On Tue, 30 Apr 2002, Seth Mos wrote:

> I think ReiserFS is still not completely NFS safe. I have lost a ReiserFS
> disk once. I am not touching it anymore.
> 
> XFS has dynamically growing inodes. The defaults are that 25% of the
> space may be allocated as inodes.
> Make sure you format it with a larger log if you would use it though. By
> default you might get a too small one for mail server IO usage.
> If you also mount the disk with more buffers you can really speed up more
> updates.
> 
> You can get a XFS enabled kernel from the http://oss.sgi.com/projects/xfs/
> site or use some the other "developers" kernels out there.
> The kernels from Andrea Arcangeli and Shawn Starr come to mind.
> 
> Cheers
> Seth
> 
> _______________________________________________
> Linux-PowerEdge mailing list
> Linux-PowerEdge at dell.com
> http://lists.us.dell.com/mailman/listinfo/linux-poweredge
> Please read the FAQ at http://lists.us.dell.com/faq or search the list archives at http://lists.us.dell.com/htdig/
> 




More information about the Linux-PowerEdge mailing list