md3000 performance issues

Barry Allard ballard at
Mon Jun 4 18:59:47 CDT 2007

Hi Christopher,

I asked our ASG engineer about this; he happens to know the lead software
developer for it.  The switch on the front of the md3000 apparently does
nothing; it's the dell el-cheapo method of recycling the md1000 chassis into
another product.  With active-active controllers, it wouldn't make sense
anyhow.  Btw, have you been able to get disk groups (LUNs) larger than 2 TB?

Topology: we have an MD3000 (15 x 300GB 10k SAS, dual ctlrs) connected to
one each 1850, 1950 and 2850 with SAS 5/e.


-----Original Message-----
From: linux-poweredge-bounces at
[mailto:linux-poweredge-bounces at] On Behalf Of Christopher McCrory
Sent: Monday, June 04, 2007 4:14 PM
To: linux-poweredge at
Subject: md3000 performance issues


  I have found very little tech info on the PV md3000 chassis. All I
found were 'point and click' type instructions.  My first tests with
iozone showed that this box screams.  When I put it into production,
read performance dropped to the floor.  My MD3000 has 8 15k SAS drives.
I tested with raid5 and raid10, both were fast with raid10 being
slightly faster.

When I went into production with raid10, three things changed.  I went
from a 2950/PCI Express controller to a 1850/PCI-X 64bit controller.  I
also changed the switch on the front of the unit to "split", assuming
that each internal controller would handle half the drives.  The third
change was within the setup GUI pairing the drives on opposite sides of
the split.  ( the test RAID10 was pairing drive 1 and drive 2, pair
drive 3 and drive 4, for production changed to pairing drive 1 and drive
11 , pair drive 2 and drive 12)

I suspect the last two changes are the problem and I am going to undo
them.  Does anyone at Dell have any performance optimization
documentation?  And/Or real world performance numbers using iozone or


Christopher McCrory
 "The guy that keeps the servers running"

To the optimist, the glass is half full.
To the pessimist, the glass is half empty.
To the engineer, the glass is twice as big as it needs to be.

Linux-PowerEdge mailing list
Linux-PowerEdge at
Please read the FAQ at

More information about the Linux-PowerEdge mailing list