• 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.

Poor M1015 SSD performance

B

brianmat

Weaksauce
Joined
Sep 1, 2011
Messages
114
I am still benchmarking my new ESX 5.1 Napp-It All-In-One and I am seeing horrible performance on my SSD drives.

My hardware is as follows:

Supermicro CSE-847E16 - 36 drive chassis
Supermicro X9DR3-LN4f+ - motherboard
Samsung ECC RAM - 192GB on the box with 48GB reserved for IO/Napp-It
M1015 flashed to 9211-IT in an x16 slot

Network configuration: All VMs are using the vmxnet3 driver.

My SATA pool is configured as follows:

Code: 
  pool: tank
 state: ONLINE
  scan: none requested
config:

	NAME                         STATE     READ WRITE CKSUM     CAP            Product
	tank                         ONLINE       0     0     0
	  mirror-0                   ONLINE       0     0     0
	    c5t5000CCA37DC4C7C5d0    ONLINE       0     0     0     2 TB           Hitachi HDS72302
	    c5t5000CCA37DC6E0E1d0    ONLINE       0     0     0     2 TB           Hitachi HDS72302
	  mirror-1                   ONLINE       0     0     0
	    c5t5000CCA37DC6E44Dd0    ONLINE       0     0     0     2 TB           Hitachi HDS72302
	    c5t5000CCA37DC72AA2d0    ONLINE       0     0     0     2 TB           Hitachi HDS72302
	logs
	  mirror-2                   ONLINE       0     0     0
	    c5t5001517972E2F74Ed0p1  ONLINE       0     0     0     23.2 GB        INTEL SSDSA2CW08
	    c5t5001517972E30212d0p1  ONLINE       0     0     0     23.2 GB        INTEL SSDSA2CW08
	cache
	  c5t5001517BB29CD4F6d0      ONLINE       0     0     0     120 GB         INTEL SSDSC2CW12

Gives the following in CrystalDiskMark:

sata.jpg


I then created a single drive pool to test the base throughput on the Intel 330 120GB drives. Configuration is as follows:

Code: 
  pool: tank1
 state: ONLINE
  scan: none requested
config:

	NAME                     STATE     READ WRITE CKSUM     CAP            Product
	tank1                    ONLINE       0     0     0
	  c5t5001517BB2A91588d0  ONLINE       0     0     0     120 GB         INTEL SSDSC2CT12

errors: No known data errors

  pool: tank2
 state: ONLINE
  scan: none requested
config:

	NAME                     STATE     READ WRITE CKSUM     CAP            Product
	tank2                    ONLINE       0     0     0
	  c5t5001517BB2ABE8D9d0  ONLINE       0     0     0     120 GB         INTEL SSDSC2CT12

I am getting this as the performance:

ssd1f.jpg


ssd2e.jpg


I still would not expect a single Intel 330 to be lower than a dual mirror of 7200 RPM drives. Especially given the benchmarks I have seen for the Intel drives this just seems way off.

I did a mirror of the Intel 330 drives and it made virtually no difference. The deltas were within a margin of error between the tests. This makes me wonder if the M1015 just isn't up to running SSDs and I will have to upgrade my controller later if I want to play in that space.

Thoughts or suggestions?
 
Last edited:
your comparing totally different things, and the benchmarks don't cover anything meaningful here.

There also isn't enough info to even guess about why any of this behaves the way it does.

Are you using nfs or iscsi? are you using sync or async writes?

Did you try comparing it to a pool of your harddrives without slog/cache ssd's?
 
This is being mounted through NFS within ESX.

Async is disabled on all 3 pools. I did try the SATA pool without SLOG and cache and there was a noticeable difference in write performance. Read performance was the same with sync=standard and sync=disabled.

I have not swapped the M1015 with a different M1015 to rule out a card issue. That's on the list for tomorrow.

Ultimately I have not been able to put together a ZFS config of 1-3 120GB SSD drives that even matches 4 7200RPM SATA drives. This is my main area for concern.

IOMeter testing shows similar differences in performance as well. I realize that benchmarks are not the be all end all of testing, but I would still expect similar tests to indicate the relative performance differences. I'm not as concerned about the raw numbers as much as I am concerned about the differences between drive types.
 
Just to follow up to my own issue, I tested last night with the ZFSGuru live CD benchmarking and it did show that the SSD drives and my planned configuration are both giving very good performance. I am not sure why the VMs aren't showing higher numbers but this weekend I am reconfiguring everything due to an ESX issue with the onboard SAS controller so I can start from scratch and see if the numbers change.

I'm also going to give iSCSI a shot just for performance testing, but I hate using iSCSI due to the LUN size limitations in ESX.
 
You must log in or register to reply here.
Back
Top