ML 350 G4 Smart Array 642 Problem & Question

[Baratin]

I have a ML 350 G4 Server with a Smart Array 642 RAID controller (Firmware 2.84).
It is running Windows Server 2003 SP2 R2.
My problem is that access to data on the server is really slow. When I said slow it means that it's really hard to work on it (it's a file server and users became really crazy waiting for minutes that the file opens).
I really thinks that the problem come from the raid controller. I've monitored the network and everything is ok, I've also checked the disks and no error are reported.
With what I read on this forum I think it's a problem with RAID controler and cache configuration.
Here is an extract of the Array Diagnostic Utility:
Cache Disabled Map Device Array Accelerator Status
Logical Drive 1 Enabled Informational

Read Cache Size 32 KB (0x0020)
Write Cache Size 0 MB (0x0000)
Disable Flag Enabled (0x00)
Cache Flags Cache Flag Status
Configured And Enabled For At Least One Logical Drive Informational
Cache Disable Code 0x0000
Total Memory Size 32 KB (0x0020)
Battery Count 0 (0x0000)
Battery Status 0x0000
Parity Read Errors 0 (0x0000)
Parity Write Errors 0 (0x0000)
Error Log All Zeroes (160 x [0x00])
Failed Batteries 0 (0x0000)
Board Present 0x01
Cache Failure Map 0x00000000
Max Error Log Entries 16 (0x10)
NVRAM Load Status 0x00
Memory Size Shift Factor 10 (0x0a)
Non Battery Backed Memory Size 32 KB (0x0020)
Memory State 0x00
Cache Auto Rev 0x00
Total Available Memory Size 64 KB (0x0040)
Percent Read Cache 100% (0x64)
Percent Write Cache 0% (0x00)
Default Percent Read Cache 0% (0x00)
Default Percent Write Cache 0% (0x00)

My questions are:
-Is there 32k not used?
-I can't change read/write cache Percent in ACU, Why?
-Is that the cause of my problems?

Thank you very much for your precious help !

Alex

PS: if anybody is interested I can give the complete Diagnostic made by RAIS Diag Utility.

[TTr]

> -Is there 32k not used?
From the quick specs in http://h18004.www1.hp.com/products/quickspecs/12069_div/12069_div.html "64 MB of DDR memory used for RAID and read cache" So it looks like 32kB is for read cache and 32kB is for RAID operations.
> -I can't change read/write cache Percent in ACU, Why?
Because you don't have the BBWC option. You can not setup any cache memory as "write" because you need the battery option.
> -Is that the cause of my problems?
Most likely. Adding the BBWC option allows you to configure write cache. I have not tried it myself but based on what I have read there is a big difference in performance.
Another factor that affects performance is the number of disks that are connected on the adapter and how they are configured.

[Baratin]

TTR thank you for your quick answer.

For your last question I have 3 HD with the following specs:
COMPAQ BF0728A4CB 67.8 GB Firmware HPB5 15000 RPM

They are configured in RAID 5.

I really don't know what to do know, the server is being replaced in 6 month but users will kill me before.

[Baratin]

Attached you will find the complete ADU report. If somebody find something wrong...

[TTr]

I don't see anything wrong in the ADU report. All 3 drives seem to be running at U320 speed. WIth 3 drives, R5 is standard practice. You could have 2 LUNs instead of 1, use 1 LUN for the OS and 1 LUN for whatever application/database is running on this server but I don't think this would make a lot of difference. You should seriously consider buying the BBWC option.

One other thing to verify is that the PCI slot that the controller is plugged in is the 133MHz slot and not the 66MHz one. (I don't know if the controller would even run at 66MHz PCI). Take a look on page 40 of this document http://bizsupport2.austin.hp.com/bc/docs/support/SupportManual/c00218063/c00218063.pdf
The fast 133MHz slot is second from the bottom on the motherboard.
The ADU calls it slot #2 but the motherboard diagram calls it bus 6.

[Baratin]

Thank you again I had a look to the server and I realized that in fact it's ML 370 G4... Sorry for that.

so I found the manual http://bizsupport2.austin.hp.com/bc/docs/support/SupportManual/c00218061/c00218061.pdf

And on page 68 it corresponds to my configuration... So the card is plugged on slot 2 (number 6) and there's nothing else in any other slot.

[TTr · Accepted solution]

That's odd. The SA642 is spec-ed as a PCI-X 133MHz card and all PCI-X slots of the ML370G4 are running at 100MHz. So apparently the SA642 can run at 100MHz as well. Unless it is having an issue because of that.
Officially the ML370G4 only supports the SA6402 array card.
Also the problema that the users are having are they also caused by the amount of processing they are doing? Maybe the server and the array card are running fine and the users are puting a lot of processing load on the server.

[Baratin]

Well you're right, we have a configuration which is unsupported by microsoft which is remote PST for outlook. But on most of our server it's ok (I'm mangaing 20 offices) This is the only one which have so much problems. But it's one of our oldest servers and it's the only one from HP the others are from DELL and Dellas array have 256kb cache so I think that's why we have this problem.

Thank you very much for your help. Know I will copy outlook .PST on all the clients and I think it will be ok.

[TTr]

I have no problem with vendor mismatch hardware either. As long as the hardware is compatible it should be fine. All I am pointing out that a 133MHz PCI card is installed in a 100MHz PCI slot. Physically that's imposible. But since it works, the card and/or the slot are adjusting at a common speed. If it is 100MHz (the card down-syncs to 100MHz) it should be fine as far as performance is concerned. If both the card and the slot are down-syncing to 66MHz perfromance will suffer. The Microsoft PST set-up is a totally different issue altogether and goes back to the server load I pointed out earlier.
If you are going to keep this server for a while, you should consider buying a BBWC kit. I don't think it is expensive, you can look in the used market (ebay) as well.

[Baratin]

Well I'll definitly try to buy the BBWC !
I've looked at the perfmon and compared to another server which is fine, I only have problems in writing operations. About read it's quitely the same but with writing ones wow !

For example Medium writing File length is 0,2 on a good server (with highest value 0,5) but n mine it's 2,8 (and highest 18 !!!) But for the Medium reading file length it's the same value.

So I'll buy it without delay and keep you informed.


Thank you very much for your help.