Sign in with
Sign up | Sign in
Your question

Why HDTach shows RAID0 slower than JBOD on 3ware 3W-7450?

Tags:
Last response: in Storage
Share
July 12, 2004 4:25:13 AM

Archived from groups: comp.sys.ibm.pc.hardware.storage (More info?)

I connected 4 units Seagate ST340016A ATA100 to a 3W-7450 (firmware &
Windows driver 7.7.0) and compared read & write speeds between RAID0 &
JBOD with HDTach 2.7 on unformatted drives, but very puzzled to find
RAID0 (4-drive striped 1024K write-cache enabled) slower than JBOD
speeds.

Here are the test results:

RAID0 Access R.Burst R.Max R.Min R.Avg W.Max W.Min W.Avg CPU%
test#1 14.0 32.9 27.4 14.5 20.6 19.5 9.2 16.0 22.5
test#2 13.9 33.0 27.8 13.8 20.7 19.6 11.9 16.0 22.8
test#3 13.9 32.9 28.3 13.5 20.4 19.2 12.8 16.1 24.0
avg 13.9 32.9 27.8 13.9 20.6 19.4 11.3 16.0 23.1

JBOD Access R.Burst R.Max R.Min R.Avg W.Max W.Min W.Avg CPU%
hdd#1 14.6 34.2 33.3 16.1 25.2 20.3 13.5 17.0 25.6
hdd#2 14.7 34.7 33.2 19.9 25.0 20.3 12.4 17.0 22.7
hdd#3 14.9 33.7 33.1 17.9 24.9 20.5 13.2 16.9 24.7
hdd#4 14.2 33.9 32.5 15.7 23.3 20.5 13.2 16.9 23.1
avg 14.6 34.1 33.0 17.4 24.6 20.4 13.1 17.0 24.0

Morever, the above JBOD speeds with the 4 drives connected to 3W-7450
are even slower than JBOD speeds with the same 4 drives connected
directly to the motherboard's ATA100 ports.

What's wrong? Please advise, thanks!
Anonymous
a b G Storage
July 12, 2004 3:24:10 PM

Archived from groups: comp.sys.ibm.pc.hardware.storage (More info?)

On 12 Jul 2004 00:25:13 -0700, hh020919@hotmail.com (hh) wrote:

>I connected 4 units Seagate ST340016A ATA100 to a 3W-7450 (firmware &
>Windows driver 7.7.0) and compared read & write speeds between RAID0 &
>JBOD with HDTach 2.7 on unformatted drives, but very puzzled to find
>RAID0 (4-drive striped 1024K write-cache enabled) slower than JBOD
>speeds.
>
>Here are the test results:
>
>RAID0 Access R.Burst R.Max R.Min R.Avg W.Max W.Min W.Avg CPU%
>test#1 14.0 32.9 27.4 14.5 20.6 19.5 9.2 16.0 22.5
>test#2 13.9 33.0 27.8 13.8 20.7 19.6 11.9 16.0 22.8
>test#3 13.9 32.9 28.3 13.5 20.4 19.2 12.8 16.1 24.0
>avg 13.9 32.9 27.8 13.9 20.6 19.4 11.3 16.0 23.1
>
>JBOD Access R.Burst R.Max R.Min R.Avg W.Max W.Min W.Avg CPU%
>hdd#1 14.6 34.2 33.3 16.1 25.2 20.3 13.5 17.0 25.6
>hdd#2 14.7 34.7 33.2 19.9 25.0 20.3 12.4 17.0 22.7
>hdd#3 14.9 33.7 33.1 17.9 24.9 20.5 13.2 16.9 24.7
>hdd#4 14.2 33.9 32.5 15.7 23.3 20.5 13.2 16.9 23.1
>avg 14.6 34.1 33.0 17.4 24.6 20.4 13.1 17.0 24.0
>
>Morever, the above JBOD speeds with the 4 drives connected to 3W-7450
>are even slower than JBOD speeds with the same 4 drives connected
>directly to the motherboard's ATA100 ports.
>
>What's wrong? Please advise, thanks!

Nothing is wrong with your system. HDTach simply gives wrong results
when you test it on Raid arrays. This is a known issue. (know to the
users, not sure if the maker has acknowledged the bug)

Marc
July 13, 2004 2:05:08 AM

Archived from groups: comp.sys.ibm.pc.hardware.storage (More info?)

On Mon, 12 Jul 2004 11:24:10 +0200, Marc de Vries
<marcdevries@geen.spam.zonnet.nl> wrote:

>On 12 Jul 2004 00:25:13 -0700, hh020919@hotmail.com (hh) wrote:
>
>>I connected 4 units Seagate ST340016A ATA100 to a 3W-7450 (firmware &
>>Windows driver 7.7.0) and compared read & write speeds between RAID0 &
>>JBOD with HDTach 2.7 on unformatted drives, but very puzzled to find
>>RAID0 (4-drive striped 1024K write-cache enabled) slower than JBOD
>>speeds.
>>
>>Here are the test results:
>>
>>RAID0 Access R.Burst R.Max R.Min R.Avg W.Max W.Min W.Avg CPU%
>>test#1 14.0 32.9 27.4 14.5 20.6 19.5 9.2 16.0 22.5
>>test#2 13.9 33.0 27.8 13.8 20.7 19.6 11.9 16.0 22.8
>>test#3 13.9 32.9 28.3 13.5 20.4 19.2 12.8 16.1 24.0
>>avg 13.9 32.9 27.8 13.9 20.6 19.4 11.3 16.0 23.1
>>
>>JBOD Access R.Burst R.Max R.Min R.Avg W.Max W.Min W.Avg CPU%
>>hdd#1 14.6 34.2 33.3 16.1 25.2 20.3 13.5 17.0 25.6
>>hdd#2 14.7 34.7 33.2 19.9 25.0 20.3 12.4 17.0 22.7
>>hdd#3 14.9 33.7 33.1 17.9 24.9 20.5 13.2 16.9 24.7
>>hdd#4 14.2 33.9 32.5 15.7 23.3 20.5 13.2 16.9 23.1
>>avg 14.6 34.1 33.0 17.4 24.6 20.4 13.1 17.0 24.0
>>
>>Morever, the above JBOD speeds with the 4 drives connected to 3W-7450
>>are even slower than JBOD speeds with the same 4 drives connected
>>directly to the motherboard's ATA100 ports.
>>
>>What's wrong? Please advise, thanks!
>
>Nothing is wrong with your system. HDTach simply gives wrong results
>when you test it on Raid arrays. This is a known issue. (know to the
>users, not sure if the maker has acknowledged the bug)
>
>Marc

What software can benchmark RAID correctly? Thanks!
Anonymous
a b G Storage
July 13, 2004 4:13:15 AM

Archived from groups: comp.sys.ibm.pc.hardware.storage (More info?)

Try HDtach 3.0 ?


On Mon, 12 Jul 2004 11:24:10 +0200, Marc de Vries
<marcdevries@geen.spam.zonnet.nl> wrote:

>On 12 Jul 2004 00:25:13 -0700, hh020919@hotmail.com (hh) wrote:
>
>>I connected 4 units Seagate ST340016A ATA100 to a 3W-7450 (firmware &
>>Windows driver 7.7.0) and compared read & write speeds between RAID0 &
>>JBOD with HDTach 2.7 on unformatted drives, but very puzzled to find
>>RAID0 (4-drive striped 1024K write-cache enabled) slower than JBOD
>>speeds.
>>
>>Here are the test results:
>>
>>RAID0 Access R.Burst R.Max R.Min R.Avg W.Max W.Min W.Avg CPU%
>>test#1 14.0 32.9 27.4 14.5 20.6 19.5 9.2 16.0 22.5
>>test#2 13.9 33.0 27.8 13.8 20.7 19.6 11.9 16.0 22.8
>>test#3 13.9 32.9 28.3 13.5 20.4 19.2 12.8 16.1 24.0
>>avg 13.9 32.9 27.8 13.9 20.6 19.4 11.3 16.0 23.1
>>
>>JBOD Access R.Burst R.Max R.Min R.Avg W.Max W.Min W.Avg CPU%
>>hdd#1 14.6 34.2 33.3 16.1 25.2 20.3 13.5 17.0 25.6
>>hdd#2 14.7 34.7 33.2 19.9 25.0 20.3 12.4 17.0 22.7
>>hdd#3 14.9 33.7 33.1 17.9 24.9 20.5 13.2 16.9 24.7
>>hdd#4 14.2 33.9 32.5 15.7 23.3 20.5 13.2 16.9 23.1
>>avg 14.6 34.1 33.0 17.4 24.6 20.4 13.1 17.0 24.0
>>
>>Morever, the above JBOD speeds with the 4 drives connected to 3W-7450
>>are even slower than JBOD speeds with the same 4 drives connected
>>directly to the motherboard's ATA100 ports.
>>
>>What's wrong? Please advise, thanks!
>
>Nothing is wrong with your system. HDTach simply gives wrong results
>when you test it on Raid arrays. This is a known issue. (know to the
>users, not sure if the maker has acknowledged the bug)
>
>Marc
July 13, 2004 5:06:06 AM

Archived from groups: comp.sys.ibm.pc.hardware.storage (More info?)

RAID0 & JBOD configurations remain unchanged, here are HDTach 3.0.1
test results:

RAID0 Access R.Burst R.Avg CPU%
test#1 14.3 45.9 30.8 10.0
test#2 14.1 46.8 30.9 6.0
test#3 14.3 45.8 30.7 6.0
avg 14.2 46.2 30.8 7.3

JBOD Access R.Burst R.Avg CPU%
hdd#1 14.8 46.3 35.2 8.0
hdd#2 14.7 46.0 33.5 0.0
hdd#3 15.0 44.7 33.6 22.0
hdd#4 14.6 43.9 31.1 0.0
avg 14.8 45.2 33.4 7.5

If such benchmarks are correct, isn't RAID0 with 3ware 7450 a total
waste of time and money?

Any other ideas? Thanks!

On Tue, 13 Jul 2004 00:13:15 +1000, OCZ Guy <ModenMr@yahoo.com.au>
wrote:

>Try HDtach 3.0 ?
>
>On Mon, 12 Jul 2004 11:24:10 +0200, Marc de Vries
><marcdevries@geen.spam.zonnet.nl> wrote:
>
>>On 12 Jul 2004 00:25:13 -0700, hh020919@hotmail.com (hh) wrote:
>>
>>>I connected 4 units Seagate ST340016A ATA100 to a 3W-7450 (firmware &
>>>Windows driver 7.7.0) and compared read & write speeds between RAID0 &
>>>JBOD with HDTach 2.7 on unformatted drives, but very puzzled to find
>>>RAID0 (4-drive striped 1024K write-cache enabled) slower than JBOD
>>>speeds.
>>>
>>>Here are the test results:
>>>
>>>RAID0 Access R.Burst R.Max R.Min R.Avg W.Max W.Min W.Avg CPU%
>>>test#1 14.0 32.9 27.4 14.5 20.6 19.5 9.2 16.0 22.5
>>>test#2 13.9 33.0 27.8 13.8 20.7 19.6 11.9 16.0 22.8
>>>test#3 13.9 32.9 28.3 13.5 20.4 19.2 12.8 16.1 24.0
>>>avg 13.9 32.9 27.8 13.9 20.6 19.4 11.3 16.0 23.1
>>>
>>>JBOD Access R.Burst R.Max R.Min R.Avg W.Max W.Min W.Avg CPU%
>>>hdd#1 14.6 34.2 33.3 16.1 25.2 20.3 13.5 17.0 25.6
>>>hdd#2 14.7 34.7 33.2 19.9 25.0 20.3 12.4 17.0 22.7
>>>hdd#3 14.9 33.7 33.1 17.9 24.9 20.5 13.2 16.9 24.7
>>>hdd#4 14.2 33.9 32.5 15.7 23.3 20.5 13.2 16.9 23.1
>>>avg 14.6 34.1 33.0 17.4 24.6 20.4 13.1 17.0 24.0
>>>
>>>Morever, the above JBOD speeds with the 4 drives connected to 3W-7450
>>>are even slower than JBOD speeds with the same 4 drives connected
>>>directly to the motherboard's ATA100 ports.
>>>
>>>What's wrong? Please advise, thanks!
>>
>>Nothing is wrong with your system. HDTach simply gives wrong results
>>when you test it on Raid arrays. This is a known issue. (know to the
>>users, not sure if the maker has acknowledged the bug)
>>
>>Marc
July 13, 2004 5:06:07 AM

Archived from groups: comp.sys.ibm.pc.hardware.storage (More info?)

I would not rely exclusively on HDTach for performance benchmarking.
What do you plan on using RAID configuration (OS / Application) for?
Suggested answers will depend on that. There is a lot of performance metrics
for which storage tuning guidlines vary a lot. Take into account size of
your data files, read/write ratio, random/sequential access ratio, block
size, average concurrent IO's. etc...
Also, you might find the following article interesting:
http://www.storagereview.com/articles/200406/20040625TC...

"hh" <hh020919@hotmail.com> wrote in message
news:2ig5f09ilquo82vumv4u78lheup2rqqptk@4ax.com...
> RAID0 & JBOD configurations remain unchanged, here are HDTach 3.0.1
> test results:
>
> RAID0 Access R.Burst R.Avg CPU%
> test#1 14.3 45.9 30.8 10.0
> test#2 14.1 46.8 30.9 6.0
> test#3 14.3 45.8 30.7 6.0
> avg 14.2 46.2 30.8 7.3
>
> JBOD Access R.Burst R.Avg CPU%
> hdd#1 14.8 46.3 35.2 8.0
> hdd#2 14.7 46.0 33.5 0.0
> hdd#3 15.0 44.7 33.6 22.0
> hdd#4 14.6 43.9 31.1 0.0
> avg 14.8 45.2 33.4 7.5
>
> If such benchmarks are correct, isn't RAID0 with 3ware 7450 a total
> waste of time and money?
>
> Any other ideas? Thanks!
>
> On Tue, 13 Jul 2004 00:13:15 +1000, OCZ Guy <ModenMr@yahoo.com.au>
> wrote:
>
> >Try HDtach 3.0 ?
> >
> >On Mon, 12 Jul 2004 11:24:10 +0200, Marc de Vries
> ><marcdevries@geen.spam.zonnet.nl> wrote:
> >
> >>On 12 Jul 2004 00:25:13 -0700, hh020919@hotmail.com (hh) wrote:
> >>
> >>>I connected 4 units Seagate ST340016A ATA100 to a 3W-7450 (firmware &
> >>>Windows driver 7.7.0) and compared read & write speeds between RAID0 &
> >>>JBOD with HDTach 2.7 on unformatted drives, but very puzzled to find
> >>>RAID0 (4-drive striped 1024K write-cache enabled) slower than JBOD
> >>>speeds.
> >>>
> >>>Here are the test results:
> >>>
> >>>RAID0 Access R.Burst R.Max R.Min R.Avg W.Max W.Min W.Avg CPU%
> >>>test#1 14.0 32.9 27.4 14.5 20.6 19.5 9.2 16.0 22.5
> >>>test#2 13.9 33.0 27.8 13.8 20.7 19.6 11.9 16.0 22.8
> >>>test#3 13.9 32.9 28.3 13.5 20.4 19.2 12.8 16.1 24.0
> >>>avg 13.9 32.9 27.8 13.9 20.6 19.4 11.3 16.0 23.1
> >>>
> >>>JBOD Access R.Burst R.Max R.Min R.Avg W.Max W.Min W.Avg CPU%
> >>>hdd#1 14.6 34.2 33.3 16.1 25.2 20.3 13.5 17.0 25.6
> >>>hdd#2 14.7 34.7 33.2 19.9 25.0 20.3 12.4 17.0 22.7
> >>>hdd#3 14.9 33.7 33.1 17.9 24.9 20.5 13.2 16.9 24.7
> >>>hdd#4 14.2 33.9 32.5 15.7 23.3 20.5 13.2 16.9 23.1
> >>>avg 14.6 34.1 33.0 17.4 24.6 20.4 13.1 17.0 24.0
> >>>
> >>>Morever, the above JBOD speeds with the 4 drives connected to 3W-7450
> >>>are even slower than JBOD speeds with the same 4 drives connected
> >>>directly to the motherboard's ATA100 ports.
> >>>
> >>>What's wrong? Please advise, thanks!
> >>
> >>Nothing is wrong with your system. HDTach simply gives wrong results
> >>when you test it on Raid arrays. This is a known issue. (know to the
> >>users, not sure if the maker has acknowledged the bug)
> >>
> >>Marc
Anonymous
a b G Storage
July 13, 2004 12:54:50 PM

Archived from groups: comp.sys.ibm.pc.hardware.storage (More info?)

On Tue, 13 Jul 2004 01:06:06 +0800, hh <hh020919@hotmail.com> wrote:

>RAID0 & JBOD configurations remain unchanged, here are HDTach 3.0.1
>test results:
>
>RAID0 Access R.Burst R.Avg CPU%
>test#1 14.3 45.9 30.8 10.0
>test#2 14.1 46.8 30.9 6.0
>test#3 14.3 45.8 30.7 6.0
>avg 14.2 46.2 30.8 7.3
>
>JBOD Access R.Burst R.Avg CPU%
>hdd#1 14.8 46.3 35.2 8.0
>hdd#2 14.7 46.0 33.5 0.0
>hdd#3 15.0 44.7 33.6 22.0
>hdd#4 14.6 43.9 31.1 0.0
>avg 14.8 45.2 33.4 7.5
>
>If such benchmarks are correct, isn't RAID0 with 3ware 7450 a total
>waste of time and money?
>
>Any other ideas? Thanks!

You could take a look at other benchmark tools used by review sites.
But really, you should just forget about HDTach completely. The
results are bogus, and I can prove that on my own system with
performance monitor.
(Tested it a few weeks ago because someone else noticed the same
problems)

But instead of benchmarks tools you can probably also get enough
information with performance monitor.

What applications will you be using the configuration for?

Marc
!