[kwlug-disc] MDADM and RAID

[John Van Ostrand]

----- "Chris Irwin" <chris at chrisirwin.ca> wrote:
> I expected a write penalty due to the parity calculations, but a
> possible read improvement.

I had an opportunity to do some testing on a system where I had to wipe the disks anyway.

It was a dual Xeon, 16GB RAM, SAS 10K 73GB drives (SEAGATE Model ST373455SS). I write 5.2GB to one drive:

        time dd if=/dev/zero of=/dev/sdb1 bs=1M count=5000

The result was 131MB/s, total time 40 seconds.

Then I created a RAID5 set with 4 of the disks and repeated. 

        md0 : active raid5 sde1[3] sdd1[2] sdc1[1] sdb1[0]
              215045760 blocks level 5, 64k chunk, algorithm 2 [4/4] [UUUU]

The result was 142MB/s, total time 37 seconds.

Those same disks as RAID0+1:

        md0 : active raid0 sde1[3] sdd1[2] sdc1[1] sdb1[0]
              286727680 blocks 64k chunks

Performed the write at 513MB/s, total time 10 seconds

I chose to do block writes to the devices on purpose since it's the most basic form of write. The test platform was CentOS 5.4.

Chris' initial tests results can be easily read improperly. Because he wisely included the "sync" in the time, we have to disregard the MB/s results that dd reports and do the math ourselves. His tests showed performance of 47MB/s for RAID5 and 78MB/s for single disk.