RAID 5 on 5 1 TB HDs - Significantly Degraded Performance

[integramodder]

I have a 2 year old Win 2008 Server box that has recently (past several months) shown serious decrease in raid disk performance. Primarily, I have seen it impacted in files transfers and file extractions (unraring). I have checked in the the web ui for the raid card that everything appears to be functioning properly. The 5 TB RAID 5 is 4 GB which translates to 3.63 GBG within windows. It currently has over 600GB free space. I have done a defrag which did not increase performance. The tests below show the performance post defrag. In a test with my system drive (2 74GB Raptors in raid 0), the system drives took 49 seconds to unzip 1.2GB while the RAID 5 took 4:39. Previously, it was rare to see it take twice as long. Any suggestions on what else I can do to troubleshoot the slow performance?

Hardware:
RocketRAID 2320 SATA Controller v1.7
WD10EADS 1TB, 7200 RPM, 32MB [QTY: 5]

Raid Settings:
Cache Policy - Write Back
Block Size - 64k
Sector Size - 512B
Read Ahead - Enabled
Write Cache - Enabled
NCQ - Enabled


Test Results

Crystal Disk:
http://i52.tinypic.com/auiqr.jpg

HD Tune Pro:
http://i56.tinypic.com/2mhv4u1.jpg

 

[Hyperlite]

Any reason (not) to suspect the raid controller?

 

[classy]

Have you tried latest drivers or card bios? What OS changes have been made? Patches maybe?

 

[integramodder]

Any reason (not) to suspect the raid controller?

I haven't considered that it would have a partial failure of working but being slow. Is this a common raid card failure?

 

[integramodder]

Have you tried latest drivers or card bios? What OS changes have been made? Patches maybe?

Checked that. It looks like they haven't updated the drivers/bios since I reformatted the server last summer.

 

[Emulex]

you have cache write back policy without a battery? that's insane dude.

if you had a battery i'd say the battery is dead and the system has gone to write-through as it should (any decent raid controller with BBWC/FBWC that detects a fault).

 

[mpx999]

Don't use RAID-5. It's too slow. Use RAID-10.

http://www.miracleas.com/BAARF/BAARF2.html

 

[Knavish]

Don't use RAID-5. It's too slow. Use RAID-10.

http://www.miracleas.com/BAARF/BAARF2.html

To slow for certain things, works fine for others.

Just remember: RAID is not a backup. It's *just* a bigger and/or faster disk.

 

[imagoon]

you have cache write back policy without a battery? that's insane dude.

if you had a battery i'd say the battery is dead and the system has gone to write-through as it should (any decent raid controller with BBWC/FBWC that detects a fault).

Maybe he is using a UPS for controlled shutdowns! (I agree no battery = bad karma)

But anyway. Remember also that disk are typically slower the farther in the spool. So as you fill the array you typically fill from fast to slow. May be that is the issue?

EDIT:

From the pic I would say the write cache is disabled for some reason. Maybe the card detected a failure of some sort? Recently had a controller fail on a SAN here and the backup controller flushed the cache to disk and the shutdown write caching and I saw numbers that that were barely better than that. (think 30MB/s write on 8 15k spools etc)