Increase Software RAID5 write speed

This is a very good post, thank you for that piece of information. I will add it to the performance tuning post.

Yeah, nice post PAPPL.

Hasta la vista, Baby!

I assume the 8192 is the amount of RAM?

The forumla is:

memory_consumed = system_page_size * nr_disks * stripe_cache_size

Stripe Cache size number is given in pages. So it is 256 pages in default setting.

The pagesize is typically 4kb, but can vary from system to system.

8192 pages equals 8192*4kb=32MB ... and this for X numbers of disks (3 +1) in my case would be then 96MB for the raid cache.... 512MB ? where do you have this number from? Or do you have that much disks in your raid?

Hey,

i already did this: http://sniptools.com/mac-osx/f…sfers-across-mac-osx-lion

And now i tried your "solution" but it did not work...
I only get about 38MB/s sent data... I don´t know hat i can do here... toooo slow for a raid 5 with 4 HDD´s over gigabit...

Edit:
At first i also got ´bout 100 to 110 MB/s transfer rates! But now, evrything i get is about 38 to 46 MB/s

am not getting anything better then what i did before running full gig network but transfer speeds of about 50-60mb

I believe that there is something else in your line. The full chain needs to be analyzed and without any further information, this is a hard challenge for us.

So there are some threads about performance tuning open with generic answers. Refer to that places and ask further questions there please.

Hi,

Is this tweak built in now in the latest OMV version?
Will it also apply to Raid10 arrays?

Thanks

Zitat von "ryecoaaron"

No, it is not built-in to OMV now.
No, it does not apply to Raid 10 because this is a striping change (raid 10 doesn't stripe).

raid 10 doesn't stripe? I think you are using a different raid 10 implementation than most of us

I have tried this advice for my Raid 5 array, but currently my Write performance is about 15-50MB/s (smaller file lower performance).

I currently have a Proliant n40L - with 4 Seagate Drives ST3000DM001-9YN166 (which are 4k format), Raid with 512k strip size.

On top of that I have LVM - then publish that over my network as a iSCSI target. (currently formated with HFS+ (osx))

My question is (after doing hours of research- and related to Raid5 write speed topic) would bypassing LVM increase writes? I have no issue with reads that currently hit 150MB/s+ (yes running on dual gig ethernet lanes).

Either that or the LVM layer is incorrectly aligned to my md device. (but then again bypassing LVM would make this issue go away...)

Cheers, in advance,
Al

Short answer:
No LVM will not decrese your write speed and also LVM is typically not misaligned.

Long answer:
LVM does not add anything to a missaligned disk. But maybe you have a lower level issue here.

Have you created partitions on your drives, bevore adding them to md?

If so, remove the partitions and create the md directly on the drives. The first case could lead to unaligned write sectors, where the later method will allways create aligned sectors. md will create the writeable area at 1mb from device start. Same is true for LVM, so that you will not have any issues with alignment.

The second thing, that could be wrong is a higher level issue - MAC, network, ethernet power, and iSCSI stacks in both OS. I am not sure how good a mac (and what mac you have) is at pure network throughput and esp. if you use iSCSI on top of it. Also where does your data come from, that you trying to write to the iSCSI drive? Is the read operation not fast enough to sustain the write performance of OMV?

So many variables in that equation. ... but LVM is not your enemy.

Thanks to the original thread starter, made a huge difference for me. OMV should do this when a RAID5 array is added

Hi there,

after every reboot, the size is again set to 256. I've already edited the rc.local ... what shall I do?

Could you post your /etc/rc.local?

cat /etc/rc.local

sure, here it is:

cat /etc/rc.local
#!/bin/sh -e
#
# rc.local
#
# This script is executed at the end of each multiuser runlevel.
# Make sure that the script will "exit 0" on success or any other
# value on error.
#
# In order to enable or disable this script just change the execution bits.
#
# By default this script does nothing.


# Tune the RAID5 configuration
echo 8192 > /sys/block/md127/md/stripe_cache_size