Hi all,

Is that combination ok for RAID 0 config? I have a
FUJITSU 36GB 10K rpm planning to get a FUJITSU
36GB 15K, should be no problem right?

TIA,
Regards,
Giovanni

Should work fine.


"Giovanni Azua" <(E-Mail Removed)> wrote in message
news:(E-Mail Removed)...
> Hi all,
>
> Is that combination ok for RAID 0 config? I have a
> FUJITSU 36GB 10K rpm planning to get a FUJITSU
> 36GB 15K, should be no problem right?
>
> TIA,
> Regards,
> Giovanni
>

In article <(E-Mail Removed)>, (E-Mail Removed)
says...
> Hi all,
>
> Is that combination ok for RAID 0 config? I have a
> FUJITSU 36GB 10K rpm planning to get a FUJITSU
> 36GB 15K, should be no problem right?

RAID doesn't really care how fast the disks are, just that they both
have the same stripe size.

--

(E-Mail Removed)
remove 999 in order to email me

As long as you accept the increased risk of data loss. If either drive
fails everything is gone...

Regards,
Hank Arnold

Giovanni Azua wrote:
> Hi all,
>
> Is that combination ok for RAID 0 config? I have a
> FUJITSU 36GB 10K rpm planning to get a FUJITSU
> 36GB 15K, should be no problem right?
>
> TIA,
> Regards,
> Giovanni
>
>

Hank Arnold wrote:
> As long as you accept the increased risk of data loss. If either drive
> fails everything is gone...
>
It should be noted that the drive speeds are indpendent of this fact,
rather it's an effect of RAID0.

I am not an expert but I would think that two drives with that much
difference in rotational speed might have a problem. Doesn't raid 0
write stripes to alternate drives? If so, wouldn't the faster drive
write or read his stripes faster and get out of sync with the slow drive
at some point and be forced to redo the read/write operation or retry to
stay in sync? I suppose that with a large buffer either on the disk or
the raid controller this might not be a problem but I'm curious.

TX John

Giovanni Azua wrote:

> Hi all,
>
> Is that combination ok for RAID 0 config? I have a
> FUJITSU 36GB 10K rpm planning to get a FUJITSU
> 36GB 15K, should be no problem right?
>
> TIA,
> Regards,
> Giovanni

On Wed, 31 May 2006 20:46:06 -0700, John Smith
<(E-Mail Removed)> wrote:

>I am not an expert but I would think that two drives with that much
>difference in rotational speed might have a problem. Doesn't raid 0
>write stripes to alternate drives? If so, wouldn't the faster drive
>write or read his stripes faster and get out of sync with the slow drive
>at some point and be forced to redo the read/write operation or retry to
>stay in sync? I suppose that with a large buffer either on the disk or
>the raid controller this might not be a problem but I'm curious.


It will be a minor mismatch in that the slower of the two is
the bottleneck, but in itself the rotational speed, or cache
size, or even (in ATA,) ATA mode doesn't necessarily matter.
The primary requirement is just that the array size has to
be accomdating to the smaller of the two drives.

Most think "ok then they have to be equal sized drives".
Usually not, it's just that it makes the most sense if
buying two new drives to make an array, but if one had an
old 100GB and 120GB HDD lying about, it could make sense to
use them and recognize that the 100GB drive capacity is the
limit, the 20GB on the other is wasted, or unutilized space.

Stripe size or stripe matching is never a concern.
Only AFTER one had assigned drives into an array, already,
does one choose stripe size. You can't choose a different
stripe size for one drive than the other because a stripe
size doesn't exist except in the array of the multiplie
drives as a logical volume already.

Speed of writing is not an issue, no matter how fast or
slow, each drive is given the I/O and the operation waits on
the drive (performance). So matching an extremely fast new
HDD with an ancient one would work fine, just be much much
slower.

The key thing about raid is the capability of the
controller. The controller itself just considers the two
drives as two plain old drives, but it's logic (or software)
rearranges how the data gets arranged onto those drives.

I have to wonder though if the OP has really thought this
out, if it is a wise choice to use RAID0. Most uses don't
need and have diminishing benefit from RAID0, particularly
on 10/15K SCSI drives, and it's a colossal waste of money if
this is all being set up on a PCI RAID card as PCI just
can't handle the throughput enough to matter... one would
be close enough to the performance at a fraction of the cost
using a single SATA WD Raptor for most uses... but then we
don't know the uses the OP has in mind.

Hi koni,

"kony" <(E-Mail Removed)> wrote in message
news:(E-Mail Removed)...
> Speed of writing is not an issue, no matter how fast or
> slow, each drive is given the I/O and the operation waits on
> the drive (performance). So matching an extremely fast new
> HDD with an ancient one would work fine, just be much much
> slower.
>
Much much slower than what? isn't it actually bound by the lower
speed disk?

> I have to wonder though if the OP has really thought this
> out, if it is a wise choice to use RAID0. Most uses don't
> need and have diminishing benefit from RAID0, particularly
> on 10/15K SCSI drives, and it's a colossal waste of money if
> this is all being set up on a PCI RAID card as PCI just
> can't handle the throughput enough to matter... one would
> be close enough to the performance at a fraction of the cost
> using a single SATA WD Raptor for most uses... but then we
> don't know the uses the OP has in mind.
>
I have the Intel E7525 MOBO that includes Adaptec HostRAID
built-in controller supporting SCSI RAID 0-1. I just did not have
the drives before but only one. Actually my setup will be:

- SATA Raptor 150GB 10Krpm 16MB cache (primary holding OS)
- RAID-0 with: 2x FUJITSU 36GB 15Krpm, 1x FUJITSU 36GB 10Krpm

I would not expect that the 10K one will slow down the other two? I expect
this RAID-0 setup to rock. Interestingly my new Raptor 10Krpm beats
by far my single SCSI FUJITSU 10Krpm using Sandra software Storage
bechmarking.

I plan to also get an external 500GB drive for periodical backups.

Thanks for your responses.

Best Regards,
Giovanni

On Sun, 11 Jun 2006 01:00:54 +0200, "Giovanni Azua"
<(E-Mail Removed)> wrote:

>Hi koni,
>
>"kony" <(E-Mail Removed)> wrote in message
>news:(E-Mail Removed)...
>> Speed of writing is not an issue, no matter how fast or
>> slow, each drive is given the I/O and the operation waits on
>> the drive (performance). So matching an extremely fast new
>> HDD with an ancient one would work fine, just be much much
>> slower.
>>
>Much much slower than what?

Than not having the very old (thus slower) drive, rather a
newer one in it's place regardless if it were identical to
the other or not.


>isn't it actually bound by the lower
>speed disk?

Define "bound"? By most definitions of "bound", no it's not
bound, the two disks still have independant performance
parameters and when both have finished their I/O, the array
as a whole has, too. So having one fast and one slow disk
can still be marginally faster than two slow ones, but
probably not much faster relative to having two newer ones
instead.

Provide specific details, otherwise there is no context.
Better yet, benchmark it yourself with the specific system
and use.


>> I have to wonder though if the OP has really thought this
>> out, if it is a wise choice to use RAID0. Most uses don't
>> need and have diminishing benefit from RAID0, particularly
>> on 10/15K SCSI drives, and it's a colossal waste of money if
>> this is all being set up on a PCI RAID card as PCI just
>> can't handle the throughput enough to matter... one would
>> be close enough to the performance at a fraction of the cost
>> using a single SATA WD Raptor for most uses... but then we
>> don't know the uses the OP has in mind.
>>
>I have the Intel E7525 MOBO that includes Adaptec HostRAID
>built-in controller supporting SCSI RAID 0-1. I just did not have
>the drives before but only one. Actually my setup will be:
>
>- SATA Raptor 150GB 10Krpm 16MB cache (primary holding OS)
>- RAID-0 with: 2x FUJITSU 36GB 15Krpm, 1x FUJITSU 36GB 10Krpm
>
>I would not expect that the 10K one will slow down the other two?

Forget that it's 10K and just focus on the actual
performance... online benchmarks would be a start if you
aren't going to benchmark it alone.