DL 580 G4 with P400 (BBWC) slow to write

[Tim Sogard]

Seeing a huge inconsistency between 2 of our Proliants.

Machine one is a DL 580 G4 with a SmartArray P400 (_with_ BBWC). Controller is set to 75% write / 25% read cache on 512MB. Raid set is 7 72GB 15K SAS drives in RAID 5. Machine has 8GB of RAM.

Machine two is a DL 320(?) with a SmartArray P400i (_no_ BBWC). Controller is 100% read and 0% write cache over 256MB. RAID set is 3 146GB 10K SATA drives in RAID 5. Machine has 4GB of RAM.

The larger (580 G4) machine performs read operations between 2 and 10 times as fast as the 320, however the 580 writes at about 90% of the speed of the 320.

Both machines are running CentOS 4.5. The DL 580 is running largeSMP, the DL 320 is running smp. Both are formatted ext3 with defaults at mount.

I have been unable to wrap my head around how the larger server is so slow at writing when it is faster at reading, has faster drives, and has such a large write cache.

I'm attaching ADU reports for both servers. (580 report attached to this post, 320 report attached to first comment).

Any help, sanity checking, or solid proof I'm crazy will be greatly appreciated.

[Tim Sogard]

Here is the 320 ADU report.

[Blazhev_1]

Hi Tim,

can you upload the .txt(and the .xml if there is such), because I can't open the reports in this format.

This is strange, since SAS is faster that SATA, further 15k vs 10k cache size and cache ratio settings...RAID 5 = more disks- more performance and so on.

The only worse thing by the dl580 g4 is that it must calculate parity for 6 drives instead of 2, but it writes simultaneousy to 6 drives at the oher hand...

Sequential or random I/O is other important aspect, since in some scenarios, cache is not used(if you have random I/O)...but even then it doesn't explain things...

In theory it should be much faster than the dl320 G5, so there seem to be something else...

Did you updated the firmware of the controller and the Hard drives using the firmware Maintenance CD 8 or 8.10...


Can you set the cache on the dl580 to 100% read or to disable it and test again? There is a similar issue with Smart array 6i and older firmware...

Please re-upload the ADUs, will check if something suspicious there

If you're crazy, at least you are not alone :)

This can

Regards,
Anton

[Tim Sogard]

Hi Anton,

Sorry if the zip got corrupted, I didn't have any problem downloading and unzipping it here, but I'm using the same platform the zip was created on.

The report files are too large to attach so I'm hosting them for now. You can view both reports (and download the xml versions) at
http://hurr.wentworthless.net/hp/


-tim

[Blazhev_1]

Hi Tim,

I took a look at the ADUs:

DL320 G5 :

P400i(FW v.2.08) with 3 SAS 10k HDDs DG146ABAB4 FW Version: HPD5 (8 Nov 2006)

DL580 G4 :

P400(FW v.2.08)
DH072ABAA6 - SAS 15k FW : HPD2....

The Firmware of the controller and the hard drives in both servers is very, very old, you should update it :

The best thing is to boot Firwmare CD and using USB key to use the HDD firmware as supplemental update(check installation instructions of HDD FW). With FW CD you will update all components at once and you avoid Firmware incompatibility :

http://h20000.www2.hp.com/bizsupport/TechSupport/SoftwareDescription.jsp?lang=en&cc=us&prodTypeId=15351&prodSeriesId=1137825&prodNameId=3288122&swEnvOID=2025&swLang=8&mode=2&taskId=135&swItem=MTX-2c5987b8362348dbb3e80d277b

P400i Firmware :

http://h20000.www2.hp.com/bizsupport/TechSupport/SoftwareDescription.jsp?lang=en&cc=us&prodTypeId=15351&prodSeriesId=1137825&swItem=MTX-3c3bfe82eb274251a092d971a3&prodNameId=3288122&swEnvOID=2025&swLang=8&taskId=135&mode=5

HDD Firmware :

http://h20000.www2.hp.com/bizsupport/TechSupport/SoftwareDescription.jsp?lang=en&cc=US&swItem=MTX-22069f1507864847b67fc3b9cd

This should increase the performance of the server.

The firmware you have is from 2006, there are a lot of fixes in the new firmware updates :

Failures that apply to your FW version :

http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?lang=en&cc=us&objectID=c01186030

http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?lang=en&cc=us&objectID=c01296286

http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?lang=en&cc=us&objectID=c01382041

If the issue persists after these updates, set the cache ratio to 100% read and test.

Hope this helps

Regards

[Blazhev_1]

hi,
did you solve the issue?