Disk performance figures on PE2650

Rechenberg, Andrew ARechenberg at shermanfinancialgroup.com
Wed Mar 26 11:03:01 CST 2003

Moving our production database system from hardware to software RAID did
not put a noticeable load on the server.  With hardware RAID the system
was averaging 95% idle; now with software RAID it's about 88% idle.
Also, since our I/O increased dramatically, it appears to the end user
as if the system is under less load and is more responsive now.

We've also had drives fail with hardware RAID and the system still
slowed down so noticeably that we had to suspend operations until the
array rebuilt.  We had to shutdown the system hard a few weeks ago due
to some other issues and all 12 RAID1 arrays were resyncing and we could
still keep the box in production (against my wishes, but I just
recommend things :) ).  It was slow, but not to the point where we had
to kick everyone off.

When I tested software RAID I made sure that it could stand up under a
real system load so I had people get on the test machine and run
processes as if it were the production box and then I failed and
re-added disks and the system was very responsive.

That being said, your mileage may vary based on your server specs and
what type of RAID cards are being used.  The test server in the above
example was a dual 2.4GHz Xeon (HT enabled) with 4GB RAM, and the
production server is a quad 1.4GHz Xeon (HT enabled) and 8GB RAM and the
hardware RAID cards were PERC3/QC's.

>> -----Original Message-----
>> From: bscott at ntisys.com [mailto:bscott at ntisys.com] 
>> Sent: Wednesday, March 26, 2003 10:42 AM
>> To: linux-poweredge at dell.com
>> Subject: RE: Disk performance figures on PE2650
>> On Wed, 26 Mar 2003, at 9:03am, 
>> ARechenberg at shermanfinancialgroup.com wrote:
>> > Basic tests (hdparm -t) for 7 disk HW v SW RAID0 (PERC3/QC v Linux 
>> > RAID) show almost a 50% increase in sequential read of 
>> 64MB file with 
>> > SW RAID
>> > (HW: 98MB/s SW: 138MB/s).
>>   Keep in mind that artificial benchmarks don't always give 
>> realistic results on an unloaded system.  If the system is 
>> busy running an actual application (like a database), and 
>> does not have to spend resources doing RAID, hardware gives 
>> you more of a benefit.  This is especially true after a disk 
>> fails and you're doing a rebuild.
>> -- 
>> Ben Scott <bscott at ntisys.com>
>> | The opinions expressed in this message are those of the 
>> author and do  
>> | | not represent the views or policy of any other person or 
>> organization. |
>> | All information is provided without warranty of any kind.  
>>             |
>> _______________________________________________
>> 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 

More information about the Linux-PowerEdge mailing list