LSI 6GB/s RAID6 Performance Issues

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

T_Minus

Build. Break. Fix. Repeat
Feb 15, 2015
7,641
2,058
113
Even with CacheCade SSD Cache the performance is terrible.

Screen clip

Be sure to scroll, and read notes at bottom it outlines what I changed between tests.

CacheCade should be worth it! I'm not sure why it's not showing CacheCade 2.0 either... any ideas?

What are proper/optimal settings for a RAID6 config? Read Ahead? Writ back? Direct IO, etc....

I'd like to use CacheCade as it seems to work great when I google, and read reviews here but wow it didn't do anything for me. Possibly due to version 1.0 vs 2.0? Not 100%.
 
Last edited:

T_Minus

Build. Break. Fix. Repeat
Feb 15, 2015
7,641
2,058
113
As you can see the S3500 appears to be operating at the Intel specs (or rather very close) based on their IOPs if memory serves me. So I'm not sure why when cached it doesn't benchmark good, other than possibly the benchmark utility doesn't make use of the cache by writing new data for each test and then reading?

I wonder if there's a different / better way to benchmark w/Cachecade/
 

T_Minus

Build. Break. Fix. Repeat
Feb 15, 2015
7,641
2,058
113
Yes, the onboard LSI 2208 cache is 1GB, tests were all done with 4GB of data as to avoid RAID CARD cache-hits only.
 

azev

Well-Known Member
Jan 18, 2013
769
251
63
Have you try enabling write back on the array ? I notice the difference between write back and write through is very significant even with cachecade on.
Anyway mine is setup with Writeback with BBu, no read ahead, direct io, and disk cache policy enabled.
I think the performance on mine is pretty good, especially when the cache is warm.
 
  • Like
Reactions: T_Minus

T_Minus

Build. Break. Fix. Repeat
Feb 15, 2015
7,641
2,058
113
Currently using/testing:

Strip Size: 64KB & 512KB tested.
(RAID6 for NAS duty, streaming videos, mp3s, etc... it has small files like "my Docs" but more for backup than linked to desktop. Not too sure optimal size if I use CacheCade + SSD though.)

Read Policy: No Read Ahead
Write Policy: Always write back.
I/O Policy: Direct IO
Disk Cache Policy: Disabled

Thank you for the input.
 

T_Minus

Build. Break. Fix. Repeat
Feb 15, 2015
7,641
2,058
113
Have you try enabling write back on the array ? I notice the difference between write back and write through is very significant even with cachecade on.
Anyway mine is setup with Writeback with BBu, no read ahead, direct io, and disk cache policy enabled.
I think the performance on mine is pretty good, especially when the cache is warm.
If you have time can you do a benchmark and also let me know which drives you're using, controller, software/hardware raid, cache, etc... :)
 

azev

Well-Known Member
Jan 18, 2013
769
251
63
here's my anvil test result, mind you the same drive is currently running about 20vm mounted via iscsi.



This is Raid6 with 14x 2Tb combination of Hitachi & Toshiba drive. Cachecade SSD is HUSSL4020ASS600... all drive connected in a supermicro chassis with sas2 expander.
 

T_Minus

Build. Break. Fix. Repeat
Feb 15, 2015
7,641
2,058
113
What about your:

Strip Size:


Read Policy:
Write Policy:
I/O Policy:
Disk Cache Policy:
(Since I had disk cache policy disabled, I wonder if that messes with CacheCade, not sure on that.)


Yours looks good with the Cachecade, and more along what I'd expect.
 

azev

Well-Known Member
Jan 18, 2013
769
251
63
Strip Size: 256K
Read Policy: No read Ahead
Write Policy: Writeback with BBU
I/O Policy: Direct IO
Disk Cache Policy: Enabled

 

ninja6o4

Member
Jul 2, 2014
92
22
8
45
Did you ever solve your RAID6 performance issues? I'm running into something similar myself and will be doing some investigating.