I have a raid 50 array on a 3ware controller. The box is running centos 6.5 and the file system is ext4.
I'm going to try some other filesystems, but could anyone suggest any alternative raid setups as well as stripe sizes I should try?
The old server uses the same controller on a centos 5.10 setup, using ext3, and it performs much faster i/o. The old 3ware setup is raid 5.
-chuck
Check if writing cache is enabled in 3ware controller configuration. You may want to have battery backup of controller RAM for that (otherwise sudden power loss will lead to loss of cache which will make a disaster on RAID level). 3ware controllers resist to enable cache in absence of battery backup or weak battery or similar.
Just a shot in a dark.
Valeri
On Wed, August 6, 2014 1:44 pm, Chuck Campbell wrote:
I have a raid 50 array on a 3ware controller. The box is running centos 6.5 and the file system is ext4.
I'm going to try some other filesystems, but could anyone suggest any alternative raid setups as well as stripe sizes I should try?
The old server uses the same controller on a centos 5.10 setup, using ext3, and it performs much faster i/o. The old 3ware setup is raid 5.
-chuck
-- ACCEL Services, Inc.| Specialists in Gravity, Magnetics | (713)993-0671 ph. | and Integrated Interpretation | (713)993-0608 fax 448 W. 19th St. #325| Since 1992 | (713)306-5794 cell Houston, TX, 77008 | Chuck Campbell | campbell@accelinc.com | President & Senior Geoscientist |
"Integration means more than having all the maps at the same scale!"
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
++++++++++++++++++++++++++++++++++++++++ Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247 ++++++++++++++++++++++++++++++++++++++++
On 6.8.2014 20:44, Chuck Campbell wrote:
I have a raid 50 array on a 3ware controller. The box is running centos 6.5 and the file system is ext4.
I'm going to try some other filesystems, but could anyone suggest any alternative raid setups as well as stripe sizes I should try?
The old server uses the same controller on a centos 5.10 setup, using ext3, and it performs much faster i/o. The old 3ware setup is raid 5.
Do you have barriers enabled? Just another shot in the dark, but 5 didn't have that. If you have battery backed Cache with your Controller, you can safely disable barriers anyway.
On 8/6/2014 5:32 PM, Markus Falb wrote:
Do you have barriers enabled? Just another shot in the dark, but 5 didn't have that. If you have battery backed Cache with your Controller, you can safely disable barriers anyway.
are you sure about this? thats not my understanding.
What I've been told is, the battery backed write cache just lets you enable writeback caching in the raid controller, write barriers at the various OS layers still have to be respected, as they ensure certain writes are completed in order.
On 7.8.2014 03:04, John R Pierce wrote:
On 8/6/2014 5:32 PM, Markus Falb wrote:
Do you have barriers enabled? Just another shot in the dark, but 5 didn't have that. If you have battery backed Cache with your Controller, you can safely disable barriers anyway.
are you sure about this? thats not my understanding.
What I've been told is, the battery backed write cache just lets you enable writeback caching in the raid controller, write barriers at the various OS layers still have to be respected, as they ensure certain writes are completed in order.
You mean there are other places besides the drive itself where I/O may be reordered? The I/O Scheduler maybe? Hmmm...
On Wed, Aug 6, 2014 at 2:44 PM, Chuck Campbell campbell@accelinc.com wrote:
I have a raid 50 array on a 3ware controller. The box is running centos 6.5 and the file system is ext4.
I'm going to try some other filesystems, but could anyone suggest any alternative raid setups as well as stripe sizes I should try?
The old server uses the same controller on a centos 5.10 setup, using ext3, and it performs much faster i/o. The old 3ware setup is raid 5.
-chuck
tuned-adm list tuned-adm profile throughput-performance
-
Chuck Campbell -
John R Pierce -
Markus Falb -
Steven Tardy -
Valeri Galtsev