[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-users] 3ware 9550SXU-4LP performance


  • To: Xen Users <Xen-users@xxxxxxxxxxxxxxxxxxx>
  • From: Michael Kress <kress@xxxxxxxxxxx>
  • Date: Fri, 20 Oct 2006 17:05:44 +0200
  • Delivery-date: Fri, 20 Oct 2006 12:49:18 -0700
  • List-id: Xen user discussion <xen-users.lists.xensource.com>

Hi,  I used xen-3.0.2-2 (2.6.16) before and the performance of my 3ware
9550SXU-4LP wasn't too bad, but now with 3.0.3.0 (2.6.16.29) throughput
decreased by about 10MB/sec in write performance.
sync; ./bonnie++ -n 0 -r 512 -s 20480 -f -b -d /mnt/blabla -u someuser

now (2.6.16.29):
Version  1.03       ------Sequential Output------ --Sequential Input-
--Random-
                    -Per Chr- --Block-- -Rewrite- -Per Chr- --Block--
--Seeks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP 
/sec %CP
matrix          20G           30546   6 26301   3           129709   3
135.2   0


before (with 2.6.16) I was at about 42MB for writing and 131MB for reading.
Have you got any ideas what I can do more for tuning?
I already did:
* StorSave: Balanced (I don't want to set it to Performance, for obvious
reasons)
* used noirqbalance parameter to prevent "nobody cared" messages related
to usb irqs
* use xfs (much faster than ext3)
* mv /lib/tls /lib/tls.disabled
   mv /lib/tls64 /lib64/tls.disabled
   ---> remarkable performance boost!(!)
* disable Queuing
* blockdev --setra 16384
* ask 3ware support
* use the newest driver for the controller (2.26.04.010)
* update firmware to FE9X 3.04.01.011
* use modules for the scsi subsys

My setup:
* Supermicro X6DH8-G2+, 4GB, 2+3.6Ghz
* use Bus Width, 64 bits, use Bus Speed 133 Mhz (I haven't noticed any
difference between 66Mhz and 133Mhz)
* 4 x Seagate ST3250820AS @3.0GBps
* Performance tests run in xen0

Why I'm writing this into the xen-users' ML: When I boot into my
distro's stock kernel (centos, i.e. 2.6.9-42.0.3), there's way bigger
performance, even with ext3, so I have to use xfs in order to reach ext3
results. So there must be some tuning that lacks here in xen.

Moreover I think, there's faster hard drives out there. I think that's
the core problem. Anyways, it's ok, performance of this machine is quite
ok. Anyways, I feel that performance in xen0 decreased whereas it
increased a bit in xenU.

Any more ideas? In other words: What the hell do the centos/redhat guys
do that the xen guys don't do. ;)
TIA, Michael

-- 
Michael Kress, kress@xxxxxxxxxxx
http://www.michael-kress.de / http://kress.net
P E N G U I N S   A R E   C O O L


_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.