• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Sanity check: RAID array performance

V

VanFanel89

2[H]4U
Joined
Apr 21, 2004
Messages
2,933
Folks,

I rebuilt my RAID array the other day and changed my ol' LSI 8408e SATA/SAS card powering 6 SATAII drives in RAID5 with an LSI 9260-i8 powering 8 drives (2 SSDs, 6 SATAIII HDDs) in a RAID1 (for SSDs) and RAID10 (for the 6 HDDs).

Strip size for the RAID10 array is 256KB, write back is enabled, HDD cache is enabled, I/O policy is cached.

The host OS is XenServer 6.2 which runs ontop of CentOS 5 and the box itself has 2x Opteron 6276's on an Asus KGPE-D16 with 48GB of DDR3 RAM - so it's quite beefy.

A quick hdparm test in xen shows average of 3200 mb/s for cached reads and between 380-400 mb/s on buffered disk reads.

Before settling on RAID10, I did a RAID6 run using the same 6 drives and had about 480-500 mb/s for buffered disk reads and about ~4,000 MB/s for cached reads.

Now the box itself will host VMs that have moderate SQL database needs, thus why I went with RAID10 - IOPS galore. But... are the read speeds really that different (and bad) between the two RAID types, or am I missing something here?
 
Your benchmark is not measuring IOPS at all but measuring large sequential performance. You should not expect better large sequential reads or writes with raid10 versus raid6.
 
Last edited:
drescherjm said:
Your benchmark is not measuring IOPS at all but measuring large sequential performance. You should not expect better large sequential reads or writes with raid10 versus raid6.
Click to expand...

This*10.

You want to measure IOPS not sequential performance.
 
Mind you, the speeds for the sequentials look about right. We are not talking enterprise SAS drives here.
 
Right - I am aware that speeds would be lower - i just wanted to double check to see if they actually are that much lower. Also yes, I did indeed expect the number not be so spectacular with consumer grade drives :p
 
Off topic: I just showed the OP to my GF to see if she could make sense of it... Took about 10secs to roll her eyes... lol :p

But yeah, you're on the right track... consumer drives makes these controllers want to kill themselves...
 
VanFanel89 said:
Now the box itself will host VMs that have moderate SQL database needs, thus why I went with RAID10 - IOPS galore. But... are the read speeds really that different (and bad) between the two RAID types, or am I missing something here?
Click to expand...

Yes the speed differences are really that different.

Here's a moderate SQL database and what it does to IO:

cLoL1zi.png


The stats you got in your tests don't really mean much because they were the only workload on your system, once you put other stuff on your drives and you take random reads into account the stats will tank all the way.

The above screenshot is from a 15+1 spindle 7.2k SATA Seagate Constellation ES 3-cell RAID50 array. Setting it to RAID10 would probably improve disk latency somewhat but it's still a matter of what else is running on the storage.
 
Got it. I actually do have the box loaded right now and it runs much better than when it had a RAID6 array... So -- I am happy :)
 
You must log in or register to reply here.
Back
Top