Need advice to get best performance

[ikosa]

Hi all, i need an advice before adding some disks and sata controller card(s) to my system.

I bought an saslp (not sas2lp) card to add more disks to my array. And i find a pci HighPoint Rocket RAID 1520 rev 2.1. (i cant find much info about unraid compability)

 

My setup details are seen on my signature. In addition to my setup in weekend i will add 2 disks to my array: ST3200542AS and ST31500341AS.

 

6 port on mobo, 8 port on saslp, (2 port on HPT1520)

How to connect my disks to get best performance?

Do i need HPT1520, can it add any performance gain, maybe for cache drive (BTW i use cache drive for just apps and downloads but i merge mkv my movies from cache to array)

 

Best choices i can think of:

Parity on mobo, 6 data drives on saslp, cache on HPT?

Parity on saslp, 6 data drives on mobo, cache on HPT?

If HPT is not compatible what should i do?

 

Thanks for any advice.

 

[tdallen]

On my setup with an older PCIex 1.1 motherboard, I found that the motherboard SATA ports were faster than my SAS2LP and therefore better for the parity and cache disks.  I eventually had to move everything onto the SAS2LP for other reasons.  I can't state that this would be true for you, but you might find it worth checking out.  The speed difference was obvious during pre-clears.

[JorgeB]

Can’t speek about the Highpoint because never used it, the SASLP works great with unraid.

 

Any configuration you choose should perform about the same for normal read / write operations to the array, the difference will be during parity checks, syncs and disk rebuilds.

 

AMDs hypertransport has a lot of bandwidth so you should have no limitations using all 6 onboard ports, because the SASLP is pcie 4x it will start limiting your parity operations with more than 4 or 5 disks.

 

So if it was me, I would leave the highpoint out for now and only add it if you need the extra ports, put the cache drive on the SASLP because it’s not used during parity operations and them start filling the onboard ports first and the SASLP next, if/when you get over 5 drives on the SASLP (not counting the cache) you can start to see some degradation during parity operations.

 

There are some users reporting high CPU load and slow parity checks with similar CPUs to yours, how is your speed now?

 

[ikosa]

Thanks for your time and advices.

I never had any issues with parity check speeds, my last parity check report:

Notice [TOWER] - Parity check finished (0 errors)

Duration: 9 hours, 22 minutes

Average speed: 89.0 MB/sec

 

I think it is reasonable. But I m not sure about Cpu load on parity check I never put an eye on it.

 

[JorgeB]

That’s a good speed and it should remain similar with up to 6 drives + cache on the SASLP

[HellDiverUK]

Highpoint cars tend to be Marvell-based with Highpoint's own firmware.  Probably not the worst performer, but they're not a LSI.

 

I have a Highpoint 2680SGL and it works just fine considering the price.  Would I buy another?  Doubtful, the LSI cards are nicer to work with.

[ikosa]

It is an old pci card with HPT372NLF chipset , I won't give any money, I just use it (if it will add any performance and if it is compatible with unraid 6) or not.

[tdallen]

A PCI card is going to slow you down - stick with motherboard and PCIex options.

[ikosa]

Thanks for everyone for the advices, i have done the upgrade.

 

These are my drive speeds before and after saslp:

root@Tower:/boot# /boot/custom/diskspeed.sh -s5 -i5

diskspeed.sh for UNRAID, version 2.2
By John Bartlett. Support board @ limetech: http://goo.gl/ysJeYV

Disk /dev/sda skipped; UNRAID boot or flash drive
/dev/sdb (Disk 4): 82 MB/sec avg
/dev/sdc (Disk 3): 93 MB/sec avg
/dev/sdd (Cache): 58 MB/sec avg
/dev/sde (Disk 2): 136 MB/sec avg
/dev/sdf (Disk 1): 104 MB/sec avg
/dev/sdg (Parity): 159 MB/sec avg

 

After adding 2 new disk to saslp:

root@Tower:~# /boot/custom/diskspeed.sh -s5 -i5

diskspeed.sh for UNRAID, version 2.2
By John Bartlett. Support board @ limetech: http://goo.gl/ysJeYV

Disk /dev/sda skipped; UNRAID boot or flash drive
/dev/sdb (Disk 4): 82 MB/sec avg
/dev/sdc (Disk 3): 92 MB/sec avg
/dev/sdd (Cache): 59 MB/sec avg
/dev/sde (Disk 2): 136 MB/sec avg
/dev/sdf (Disk 1): 104 MB/sec avg
/dev/sdg (Parity): 157 MB/sec avg
/dev/sdh: 97 MB/sec avg
/dev/sdi: 89 MB/sec avg

 

After swapping my slovest disks to saslp and adding them to array (cache and disk4):

root@Tower:~# /boot/custom/diskspeed.sh -s5 -i5

diskspeed.sh for UNRAID, version 2.2
By John Bartlett. Support board @ limetech: http://goo.gl/ysJeYV

Disk /dev/sda skipped; UNRAID boot or flash drive
/dev/sdb (Disk 5): 97 MB/sec avg
/dev/sdc (Disk 3): 92 MB/sec avg
/dev/sdd (Disk 6): 89 MB/sec avg
/dev/sde (Disk 2): 136 MB/sec avg
/dev/sdf (Disk 1): 104 MB/sec avg
/dev/sdg (Parity): 157 MB/sec avg
/dev/sdh (Disk 4): 82 MB/sec avg
/dev/sdi (Cache): 52 MB/sec avg

 

Generally speeds are pretty same except cache. Which is probably because some dockers are using the disk at that moment.