Re: Performance of a software raid 5

[Johannes Segitz <johannes.segitz@gmail.com>]

On Tue, Apr 21, 2009 at 7:46 AM, Neil Brown <neilb@suse.de> wrote:
> I suspect you will see that improve when you add another drive that it
> isn't running degraded.

Well, i didn't but it doesn't really matter since the crypto layer slows
it down to a crawl

Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5]
[raid4] [raid10]
md6 : active raid5 sdd1[0] sdc1[1] md7[3] md8[4] sde1[2]
      3907039232 blocks level 5, 256k chunk, algorithm 2 [5/5] [UUUUU]
      bitmap: 1/15 pages [4KB], 32768KB chunk

md7 : active raid0 sdg1[0] sda1[1]
      976767744 blocks 128k chunks

md8 : active raid0 sdh1[1] sdb1[0]
      976767744 blocks 128k chunks

So everything is okay now, no missing drive and the bad drive is now gone.

for d in /dev/sd[cde]1 /dev/md[678] /dev/mapper/daten bigfile
do
   echo $d
   dd if=$d of=/dev/null bs=1M count=1000
done

/dev/sdc1
1048576000 bytes (1.0 GB) copied, 6.13302 s, 171 MB/s
/dev/sdd1
1048576000 bytes (1.0 GB) copied, 12.2261 s, 85.8 MB/s
/dev/sde1
1048576000 bytes (1.0 GB) copied, 11.8026 s, 88.8 MB/s
/dev/md6
1048576000 bytes (1.0 GB) copied, 6.42977 s, 163 MB/s
/dev/md7
1048576000 bytes (1.0 GB) copied, 9.51655 s, 110 MB/s
/dev/md8
1048576000 bytes (1.0 GB) copied, 7.97321 s, 132 MB/s
/dev/mapper/daten
1048576000 bytes (1.0 GB) copied, 28.6309 s, 36.6 MB/s
bigfile
1048576000 bytes (1.0 GB) copied, 31.9715 s, 32.8 MB/s

So the raid works okay, although i'm not thrilled by 163 MB/s
read speed when i see what the underlying devices are capable
of. But the real bad drop seems to be the crypto and with
http://tynne.de/linux-crypto-speed in mind it seems quite a
reasonable speed, although i'm disappointed since i expected
more. Next time i won't expect but do better tests beforehand.

Thanks everyone for their help.
Johannes