topic Re: How to know if BBWC is enabled in ProLiant Servers (ML,DL,SL)

How to know if BBWC is enabled

Edson Marquezani Filho — Mon, 17 May 2010 12:14:01 GMT

Hello, everybody.

I've been working with some HP Servers (ML and DL) for a while, but I've never gone much deep on hardware details or related stuff.

Sometime ago I experienced very bad write performance issues with an e200i controller on a RAID 5 setup, but I left the problem unsolved once I guessed it wouldn't be a serious problem on that specific case, and I wasn't willing to buy more hardware.

Anyway, now I decided to go throught that matter, in order to choose a correct hardware setup so that I can have fault tolerance and (relatively) good performance without having an actual dedicated Storage.

All right, after some reading I have figured out that the whole thing about write performance on HP servers controllers depends on the write cache. So, I became aware of the BBWC, and the need of a battery module to have such cache. (No, I didn't have a cache battery at that time, so I suppose that was the origin of my problem.)

Ok, now I have a brand new DL180 on my hands, which came to me with BBWC (let's call) "hardware requisites" (cache memory and battery modules) already installed on it. So, I discovered that HP CLU CLI tool and installed it.

What I want to know is: how can I know wheter BBWC is actualy enabled on the controller or not? There's no explicit reference of it on controller details.
Would that be called "Array Accelerator"? I guess that "Drive Write Cache" reffers to disk internal caches, not the BBWC, am I right? Is that BBWC enabled automatically when you upgrade your controller with battery and memory modules?

And a last question: Would be that a good choice for fault tolerance, a minimun of scability and reliability, without having a much big drawback on performance, on what regards storage?

Can you help me to understand those things?

Thank you very much.

Re: How to know if BBWC is enabled

TTr — Mon, 17 May 2010 13:40:17 GMT

Yes the BBWC info is under the array accelerator status section of the ACU/ADU report. The look of the report changes with controllers/firmware/ACU versions. The cache memory, read/write allocations, battery count, battery status are all listed under this section.

Drive Write cache refers to the drive internal cache, that should stay disabled when you use the drives on a raid controller.

The smart array controllers offer a good disk based fault tolerance. Any of the raid1/10/5 are good redundant raid types. You should leave disk to be used as hot spare and monitor the controller status for any events.

Re: How to know if BBWC is enabled

Terry Hutchings — Mon, 17 May 2010 13:59:28 GMT

One easy way to tell if you have a battery backed cache is to see what read/write ratio is currently set. The default ratio has been 50/50 for a long time, but I believe this has recently been changed to 25/75 as the default. This should be viewable under the controller settings within the ACU and ADU.

RE your choice for fault tolerance:
Every selection has a tradeoff. Raid 5 has poorer write performance, but better read performance. Raid 1 has better write performance, but poorer write performance. If you can afford the extra disks raid 1+0 (10), with 4 or more drives, would be the the best option for performance.

Re: How to know if BBWC is enabled

Edson Marquezani Filho — Mon, 17 May 2010 14:12:24 GMT

I'm reviewing my servers right now, installing "hpacucli" tool and retrieving information about their controllers. I intend to performance I/O disk tests within this week, to have more data to help me understand performance, specially when it comes to RAID5.

I'll come soon to report what I've got and discuss a little more about that matter with you.

Thank you for the answers, for now.

Re: How to know if BBWC is enabled

Erdogan Temur — Tue, 18 May 2010 05:19:15 GMT

Hi Edson,

See the attached manual.
BBWC enabler can be a problem.

Regards.

Re: How to know if BBWC is enabled

Edson Marquezani Filho — Mon, 24 May 2010 10:12:12 GMT

Well, let's see if you can help me to understand better a case I had.

Sometime ago I deployed a ML 350 G5 as a file server, with 4 SATAII disks set, arranged in a RAID5 volume. The controller was an E200i with 128MB of cache, and BBWC enabled at 50%/50%, as I understand now and you can see below:

Smart Array E200i in Slot 0 (Embedded)
Bus Interface: PCI
Slot: 0
Serial Number: QT8CMP7964
Cache Serial Number: P9A3A0B9SX01VT
RAID 6 (ADG) Status: Disabled
Controller Status: OK
Chassis Slot:
Hardware Revision: Rev A
Firmware Version: 1.82
Rebuild Priority: Medium
Expand Priority: Medium
Surface Scan Delay: 15 secs
Post Prompt Timeout: 0 secs
Cache Board Present: True
Cache Status: OK
Accelerator Ratio: 50% Read / 50% Write
Drive Write Cache: Disabled
Total Cache Size: 128 MB
No-Battery Write Cache: Disabled
Cache Backup Power Source: Batteries
Battery/Capacitor Count: 1
Battery/Capacitor Status: OK
SATA NCQ Supported: False

By that time, I performed some disk performance tests with IOZone, and the results I got were not as I expected (I mean worse). I tried all chunk sizes for RAID, from 8k to 256k, but the rates stayed around the same values for all of them. I followed RAID/LVM/Partition/EXT3 alignment recomendations too. You can see an example of my results here[1].

As you can see, for files a little large (bigger than 100MB), write rates goes from 50 to 30 MB/s, basiclly. I thought it very bad.

Should I expect the same rates for a P410/P410i (or P212) with at least 256MB of cache and BBWC enabled, or can I be more optmistic about it?

[1] http://pastebin.com/20vGpRxD