topic Re: Smart Array 5304 Peformance considerations in Disk Enclosures

Smart Array 5304 Peformance considerations

Jörg-Stefan Sell — Mon, 23 Jan 2006 14:16:33 GMT

I have a Raid 5 with 10 hard drives in a 4300 enclosure attached to a Smart Array 5304 with 128MB of cache.

Here's my question: Will I see a performance increase of the logical drive if I arrange the drives in two different 4300s, each attached to a different port on the controller?

Thanks in advance,
Jörg-Stefan

Re: Smart Array 5304 Peformance considerations

e4services — Tue, 24 Jan 2006 10:56:23 GMT

sometimes.
Performance is all relative.
Usage, application, etc. Adding a cahe module is the most dramatic basic performance increase, giving some breathing room for writing the RAID.

Re: Smart Array 5304 Peformance considerations

Jörg-Stefan Sell — Fri, 27 Jan 2006 15:29:08 GMT

I finaly had the time to give it a try:

Before, a single threaded Disk I/O tool showed a 45MB/sec write rate and a 72MB/sec read rate; after moving half of the drives to a second enclosure attached to port 3 the read rate increased to 85MB/sec, write performance did not show any change.

Re: Smart Array 5304 Peformance considerations

Steven Clementi — Fri, 27 Jan 2006 22:23:58 GMT

JÃ¶rg:

Thats sounds about right. You should expect to see a read performance since you are now utilizing more of the card's buses instead of taxing a single bus.

As for Writing, I would expect that you would see a bigger difference on a more intensive test, but simple writes would probably not show much change.

Can you perform another test assuming you have the time, this time try copying or moving a large amount of data. If you did this already in the initial test, then just disregard.

Steven

Re: Smart Array 5304 Peformance considerations

Jörg-Stefan Sell — Sat, 28 Jan 2006 08:10:15 GMT

Thanks for your response, Steven.

What do you consider a larg file? I did my initial tests with 500MB files...

I personaly would have expected that the 33Mhz slot in which the 5304 is located is a bottleneck, it's obviously wrong, I would say.

What do you think, will I see an additional performance increase if I move the controller to a machine with a 66Mhz PCI slot?

What concerns me a bit: During disk I/O the system time shown by Windows Task manager (W2K3+SP1) significantly increases, about 25% on a four way system. Firmware and drivers are up to date... Is this a normal, expected behaviour?

Is there any documentation from HP which covers this disk I/O performance topic in more detail?

Re: Smart Array 5304 Peformance considerations

Jacob Zeeman — Thu, 09 Feb 2006 11:49:39 GMT

Hi Jorg,

I don't want to disagree with any of the previous replies. However if you plan on spanning a single logical drive/array onto two different SA4300 controllers you will likely not be able to access your storage.

This is because you are unable to span an array to mulitple controllers, you can only span channels. With the SA5304 all four ports are on the same controller.

Just wanted to give you a warning before you attempt this process.

(Sorry if someone already covered this. I didn't look too thoroughly at the previous posts.)

Jake Zeeman
HP Server Storage

Re: Smart Array 5304 Peformance considerations

Jörg-Stefan Sell — Thu, 09 Feb 2006 13:03:32 GMT

Don't worry, the scenario is one controller with multiple enclosures, not the other way.

But many thanks for your attention!