Redundant NFS storage setup (part 3) : The disappointing PERC5/E

[Matthias Saou]

Tino Schwarze wrote :

> On Thu, Jan 03, 2008 at 02:16:06PM +0100, Matthias Saou wrote:
> 
> > > Are you using the latest megaraid_sas? I've seen speed improvements
> > > using the new driver.
> > 
> > Yes, on RHEL5.1 x86_64 rebuilt with dkms from the Dell provided
> > sources :
> > 
> > # modinfo megaraid_sas
> > filename:       /lib/modules/2.6.18-53.1.4.el5/extra/megaraid_sas.ko
> > description:    LSI Logic MegaRAID SAS Driver
> > author:         megaraidlinux at lsi.com
> > version:        00.00.03.16
> > [...]
> > 
> > On top of the RAID array, I use XFS :
> > # modinfo xfs
> > filename:       /lib/modules/2.6.18-53.1.4.el5/extra/xfs.ko
> > license:        GPL
> > description:    SGI XFS with ACLs, security attributes, realtime, large
> > block/inode numbers, no debug enabled
> > author:         Silicon Graphics, Inc.
> > srcversion:     BBB854051112D7CF26B397E
> > depends:        
> > vermagic:       2.6.18-53.1.4.el5 SMP mod_unload gcc-4.1
> > 
> > I always create the filesystem with :
> > # mkfs.xfs -f -L data -i size=512 /dev/sdb1
> 
> Have you tried specifying the RAID stripe size and width -d su=128k,sw=x?

I hadn't, no, and maybe XFS optimizes the software RAID array on its
own as it can easily access the array's details... but I just
tried with -d su=64k,sw=15 (as I created the PERC's RAID5 on the 15
disks), but it didn't improve anything

Maybe tweaking the XFS filesystem settings could help a little, but I
really don't think that's where the underlying problem comes from.
Running this command twice in parallel gives me between 35 and 50MB/s
read on the block device :

cat /dev/sdb >/dev/null

Again, the array is still busy with its background initialization, so
I'll try again once it's finished, but I doubt I'll see any major
improvement.

Matthias

-- 
Clean custom Red Hat Linux rpm packages : http://freshrpms.net/
Fedora release 8 (Werewolf) - Linux kernel 2.6.23.9-85.fc8
Load : 0.82 1.18 1.20