RAID question on initial setup and testing

[dariushou]

I'm fixing to do some benchmarks on my new RAID Array, but have a couple of questions.

1. When i run the benchmarks do i need to have anything (files, etc..) on the drives i'm testing? I would think so, otherwise, what would the benchmark be testing??

2. Also, my ARRAY shows up in my computer on XP as one drive as it should, but is this the only way to know if you setup it up right. I went into my RAID bios (Adaptec 3400s) and it says it's setup. The reason i guess i ask is that it didn't take anytime at all to do --just seemed to easy). Formatting NTFS was the only thing that took some time.

 

[dariushou2]

Please help with the first question as i really need it.

Thanks ahead of time.

 

[lhgpoobaa]

You dont need files on a disk to benchmark it.
Usually the benchmark program uses its own temp files. And with certain benchmark programs actually give bad results when the drive contains files.
The drive has to be formatted though.


<b><i>The Very Hungry Caterpillar</i> - George W. Bush's favorite childhood book.
Note: This book was first published a year after Mr Bush graduated from College.</b>

 

[dariushou2]

I thought that the performance (stripesize/clustersize) of RAID depends on the size of files you use. How does the benchmarking program know what size i use? For example, a 64K stripesize is much better than 8K when dealing with video files. I'm really confused here. Help would be greatly appreciated.

 

[lhgpoobaa]

You want help? Read the RAID FAQ... More comprehensive than any reply i could write!

<b><i>The Very Hungry Caterpillar</i> - George W. Bush's favorite childhood book.
Note: This book was first published a year after Mr Bush graduated from College.</b>

 

[dariushou2]

It doesn't address this issue.

 

[lhgpoobaa]

Sighs.. im pretty sure its in there.

If you are going to do long sustained reads and writes then you will probably benefit from large stripe sizes.

Dealing with lots of small interactions would be better with a small stripe size (i think)

its really down to personal prefrence.

<b><i>The Very Hungry Caterpillar</i> - George W. Bush's favorite childhood book.
Note: This book was first published a year after Mr Bush graduated from College.</b>

 

[dariushou]

Yes, it does address small versus large stripesizes, but it does not explain why you would not want files on the hard drives you are testing during a benchmark. I would think that it is imperative that files be on the hard drive so that the benchmark would see what size of files you use in order to do a proper benchmark. Maybe i'm wrong--just need some clarification.


Thanks Again

 

[lhgpoobaa]

Ive said before, the benchamrk programs make their own files, or simulate doing so. So for good HDD benchamrk programs the degree of fragmentation and how full the drive is has no bearing on the result.
Poor benchmark programs, such as sandra actually give lower results the fuller ones drive is as the existing files get in the way.

What i suggest if you go to <A HREF="http://www.storagereview.com" target="_new">http://www.storagereview.com</A> and read their testing metodology.

<b><i>The Very Hungry Caterpillar</i> - George W. Bush's favorite childhood book.
Note: This book was first published a year after Mr Bush graduated from College.</b>