Drop in Performance - HP P822

5teve

Active Member
Jan 23, 2015
106
35
28
45
Perth, Australia
Hi All

Looking for some expert opinions..

I have a HP P822 hooked up to 12 hitachi Ultrastars, while testing the system prior to using it full time I was getting some very fast benchmarks with 8x disks running (I was waiting for some 8088 to 8088 cables to arrive for the external to internal adapter)

Benchmark with 8x disks and 'not much data' on the disks



Now benching the 12 drive single volume with 10% space taken up - I had to reformat the drive in 16k cluster size as it was over the size limit for the 4k ntfs default - not sure if this would affect anything -



As you can see the read speed is very low and the write speed is about the same as my 8x 500gb consumer hitachi drive array connected to a perc 5

Any ideas before I move all the data again destroy the array, recreate, create the volume, benchmark, add data, rebenchmark..?? Hard to do all that while trying to run a business off the same machine :)

Steve
 
Last edited:

Patriot

Moderator
Apr 18, 2011
1,306
690
113
" not sure if this would affect anything "
Going to go ahead and say.... probably did.
Make sure you adjusted your stripe size for increase in disk count as well.
 

5teve

Active Member
Jan 23, 2015
106
35
28
45
Perth, Australia
Thanks Patriot... :)

So what are the recommendations for stripe size / cluster size etc for the HP?

I currently have

Status OK
Parity Initialization Status Initialization Completed
Drive Type Data
Size 18.19 TiB (20.00 TB)
RAID Level RAID 6 (ADG)
Legacy Disk Geometry (C/H/S) 65535 / 255 / 32
Strip Size / Full Stripe Size 256 KiB / 2560 KiB
Disk Name \\.\PhysicalDrive0 (Disk0)
Disk Partition Information Partition Number: 3, Size: 18.1 TiB, Mount Point: E:\

And I formatted NTFS with 16k clusters (something to do with vss)

Any help will be appreciated.

Oh and just out of interest.. I am just running defrag and for fun ran the benchmark again.. with considerably higher results.. while defrag was running .. confused..

Steve
 

5teve

Active Member
Jan 23, 2015
106
35
28
45
Perth, Australia
Ok I have been trying to teach myself in a very limited sense, as I only have limited time available currently.

I'm just looking at partition alignments.. apparently they are aligned, but partition 1 (reserved) is offset 17kb and partition 2 (primary) is offset 129mb

Do these figures look right?

using wmic i get the following - the raid array is disk 0 but doesnt show the reserved partition..

 

Patriot

Moderator
Apr 18, 2011
1,306
690
113
I don't know... I run those suckers on linux... Do you happen to have the latest driver or were you just using the inbox driver?
 

5teve

Active Member
Jan 23, 2015
106
35
28
45
Perth, Australia
I'm running the HP 63.8.0.64 driver, which I think is the latest, and firmware 6.34

the strange thing is I did a full defrag (via auslogig disk defrag) and the benchmark speed came back.. but then this morning.. it dropped off again.. so i suspect something is interfering with the system

I'm still trying to learn about stripes / chunks / full stripe size / clusters ect... then i'll make a copy of the folders to other drives wipe the array and start from scratch.

Steve
 
  • Like
Reactions: Patriot

mrkrad

Well-Known Member
Oct 13, 2012
1,244
52
48
Do you have the sector patrol at the default of 3 seconds? might bump that up higher, and make sure power savings mode on the bios is set to FULL PERFORMANCE!
 

5teve

Active Member
Jan 23, 2015
106
35
28
45
Perth, Australia
Thanks Mrkrad

The surface scan analysis priority was set to 3, its now the max of 30, but turning off made no difference. When you say the bios are you referring to the motherboard bios or the HP card bios?

Another test i have just done is to change the cache to 100% read.. and apart from the writes going through the floor.. the reads were still very poor.. i'm wondering whether there may be a disk or cable causing a slow down? when I eventually rebuild the array i'll test the original 8 disks, then add 1 disk at a time to see what happens.

Thanks

Steve
 

5teve

Active Member
Jan 23, 2015
106
35
28
45
Perth, Australia
OK I have done some more work on this.. it appears that crystal disk mark screws with the drives. If I ran a benchmark using crystal, and then ran atto.. the reads dropped right off. seems to take a while or a significant event (such as delete a heap of files) and the read speeds come back again. To eliminate any config issues or anything I have now reset the controller, rebuilt the array and restoring files (while parity is initialising) before I restored the files I benched it using atto using various strip sizes - as follows but not too extreme from the standard 256kb

128kb strip - 16k allocation unit (ntfs)



256kb strip - 16k allocation unit (ntfs)



512kb strip - 16k allocation unit (ntfs)



I'm sure there are variables in there that cause variation, especially as it is initialising parity while doing the benchmark.

Not sure also what causes the drop after 512kb.. i'm sure one of you knowledgeable guys will know tho :)

Once data is back on i'll bench it again (i'm on 256kb strip) and then try my crystal disk theory...

Steve