needee

raid10 is faster. but in my case its slower.
please help me, what I am doing wrong.

/dev/md0 is raid1 consist upon sda7, and sda8.
/dev/md1 is raid1 consist upon sda9, and sda10.

/dev/md2 is raid0 consist upon md0, and md1.


# mdadm --create /dev/md2 --chunk=64 --level=0 --raid-devices=2 /dev/md[01]

# cat /proc/mdstat
Personalities : [raid1] [raid0]
md2 : active raid0 md1[1] md0[0]
626176 blocks 64k chunks

md1 : active raid1 sda10[1] sda9[0]
313152 blocks [2/2] [UU]

md0 : active raid1 sda8[1] sda7[0]
313152 blocks [2/2] [UU]

unused devices: <none>

performed the following test that proves that RAID10 is slower.

md2 is raid10
# hdparm -t /dev/md2

/dev/md2:
Timing buffered disk reads: 86 MB in 3.07 seconds = 27.97 MB/sec

md0 and md1 are raid1
# hdparm -t /dev/md0

/dev/md0:
Timing buffered disk reads: 180 MB in 3.00 seconds = 59.95 MB/sec

please help/suggest.
Regards

coopstah13

Why don't you create raid device using --level=10 and pass in the 4 disks rather than manually create 2 raid 1's, then raid 0 them? I'm not sure if it matters or not, but its definitely something to try.

needee

>Why don't you create raid device using --level=10 and pass in the 4 disks rather than manually create 2 raid 1's, then raid 0 them

Two reasons

1, actually I followed the following tutorial
The Linux Juggernaut: How to implement RAID10?

2, and as per the definiton of RAID10
"RAID-10 can be implemented as a stripe of RAID-1 pairs. For example, given 6 devices, you may configure them as three RAID-1s A, B and C, and then configure a RAID-0 of ABC."
source: Linux RAID best practices - Strugglers

thats why I think, using --level=10 and passing the 4 disk is not good/safe/reliable... but I know I am not an expert.

So, I will try --level=10.
does --level=10 gives redundancy/fault-tolerance ?

Thanks,
Regards
needee

Irithori

Your raid 10 is built by 4 partitions of the SAME disk.
This has only educational value.

In practice, this is more than useless.
You wont gain any data protection in case of disk failure
and OF COURSE this is slower.

That one poor disk has to write each request 4 times..
Well, to be more precise:
A write request by an application is first split by the raid0.
Then each of the raid1s has to write its part twice.


Other than that, coopstah13 is right.
Don't build the raid1s and the raid0 yourself, but rely on mdadm to get that sorted for you via raid10.
Use 4 individual drives, and a halfway decent CPU, then see your IO performance fly

coopstah13

Good catch on the partitions all on one disk, I didn't even see it. That will of course be significantly slower!

needee

coopstah13 and Irithori, thanks for your nice help/support.

I used --level=10, and its working superb.

by the way, then why these tutorial exists
The Linux Juggernaut: How to implement RAID10?

Irithori

Because not everything on the internet is the absolute truth.
Shocking revelation, I know