HELP! Problems benchmarking a raid pair

[Gordon]

I have a pair of identical Maxtor 160GB IDE hard drives, each on its own
channel of a Promise ATA100 PCI controller (not RAID). I am running Windows
XP Pro and have configured them as a single striped RAID volume using the
Windows XP partitioning option. My boot drive is a single drive on IDE0. I
have tried to benchmark the new striped volume in order to see just how much
faster (roughly) the raid is than the boot drive, but every attempt to do so
has failed.

As I understand it, a drive IO benchmark doesn't measure the performance of
the raid pair, so I need to do a file system benchmark. So far I have tried
Sandra (it hangs), and Iozone which fails. I have 512MB in my machine and
Izone fails when testing the 512MB file size... although it measures the
boot drive fine. I have "memtested" my memory to death and I'm pretty sure
my memory is fine. I've tried copying multi-gig zip files to and from the
RAID pair and then extracting them to test their integrity... all is fine.

What's up with these benchmarking tools?

 

[Rod Speed]

I have a pair of identical Maxtor 160GB IDE hard drives, each
on its own channel of a Promise ATA100 PCI controller (not RAID).
I am running Windows XP Pro and have configured them as a single
striped RAID volume using the Windows XP partitioning option. My
boot drive is a single drive on IDE0. I have tried to benchmark the
new striped volume in order to see just how much faster (roughly)
the raid is than the boot drive, but every attempt to do so has failed.

As I understand it, a drive IO benchmark doesn't measure the performance
of the raid pair, so I need to do a file system benchmark. So far I have tried
Sandra (it hangs), and Iozone which fails. I have 512MB in my machine and
Izone fails when testing the 512MB file size... although it measures the
boot drive fine. I have "memtested" my memory to death and I'm pretty sure
my memory is fine. I've tried copying multi-gig zip files to and from the
RAID pair and then extracting them to test their integrity... all is fine.

What's up with these benchmarking tools?

Neither are particularly useful when benchmarking raids

Have a look at what the reviews use and use those.
http://www.xbitlabs.com/articles/storage/

That also lets you see what other systems can do too.

 

[Folkert Rienstra]

I have a pair of identical Maxtor 160GB IDE hard drives, each on its own
channel of a Promise ATA100 PCI controller (not RAID). I am running Windows
XP Pro and have configured them as a single striped RAID volume using the
Windows XP partitioning option. My boot drive is a single drive on IDE0. I
have tried to benchmark the new striped volume in order to see just how much
faster (roughly) the raid is than the boot drive, but every attempt to do so
has failed.

As I understand it, a drive IO benchmark doesn't measure the performance of
the raid pair,

It does, however HD Tach only works on drives under bios control.

so I need to do a file system benchmark. So far I have tried Sandra
(it hangs), and Iozone which fails. I have 512MB in my machine and
Izone fails when testing the 512MB file size... although it measures
the boot drive fine.

Try ATTO bench, http://www.attotech.com
I think it is under ATTO ExpressPro-Tools

 

[Gordon]

Try ATTO bench, http://www.attotech.com
I think it is under ATTO ExpressPro-Tools

Thanks. That one worked as long as I didn't choose "direct i/o". I have to
say that I was disappointed with the results. The single drive seems to be
faster than the raid pair. I guess only hardware raid is worth it?

 

[Folkert Rienstra]

Thanks. That one worked as long as I didn't choose "direct i/o".

Unfortunately that is the one you want to use to get HD Tach comparable figures.

I have to say that I was disappointed with the results.
The single drive seems to be faster than the raid pair.

I wouldn't trust that number. On my (single) drive that gives
me numbers that are double what the drive actually can do.

I guess only hardware raid is worth it?

Unlikely and too soon to say.
If you can't find a benchmark to do the comparison job for you then you
may have to revert to timing the big file copy with stopwatch method.

 

[Gordon]

Unfortunately that is the one you want to use to get HD Tach comparable figures.

I wouldn't trust that number. On my (single) drive that gives
me numbers that are double what the drive actually can do.


Unlikely and too soon to say.
If you can't find a benchmark to do the comparison job for you then you
may have to revert to timing the big file copy with stopwatch method.

The problem is, where do I copy it from so that the measurement is
independent of the read speed of the source?

 

[Folkert Rienstra]

The problem is, where do I copy it from so that the measurement is
independent of the read speed of the source?

Ramdisk?