Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Performance Gains Using 10K or 15K rpm drives

2 views
Skip to first unread message

Walter L. Preuninger II

unread,
Nov 22, 2001, 10:32:39 AM11/22/01
to
I am looking at upgrading my drive packs on a Mylex DAC960PL raid
controller(see previous post). The controller from what I can tell is
fast/wide. All the drives that I am finding are Ultra160.

My question is, will I see any peformance gains using 10K or 15K rpm drives
over 7200 rpm drives?

Currently I have 4 ST318404LW for my Oracle Database and 4 ST39173W for data
storage. My plain is to move the 18Giggers to the data storage and install
36 or 70+ for the database. Backup is not a problem, I am getting a 10 slot
LTO tape changer.

Thanks In Advance,

Walter


David A. Lethe

unread,
Nov 22, 2001, 11:10:25 AM11/22/01
to

Performance gains are relative to the percentage of time your disks
are doing seeks and transfers. You need to first establish how many
and what type of I/Os and latency you are currently seeing, then
factor in the performance gains for those parameters with the new
disks.

In simplistic terms, if you are only doing I/O 20% of the time, then
you won't see any advantage. If you are 90% I/O bound, then you will
see significant advantage.


daytripper

unread,
Nov 22, 2001, 11:52:58 AM11/22/01
to
On Thu, 22 Nov 2001 15:32:39 GMT, "Walter L. Preuninger II"
<taz...@hotmail.com> wrote:
>
>My question is, will I see any peformance gains using 10K or 15K rpm drives
>over 7200 rpm drives?

Given that database engines are usually tied to synchronous disk IO (for write
commits, at least) it's a dead certainty you'll see improved performance with
faster drives (all else being equal).

How much would be a tougher question to answer ;-)

/daytripper

Bodo Mueller

unread,
Nov 22, 2001, 1:09:55 PM11/22/01
to
Hello

As the others already pointed out, it depends on your application.

you would see no preformance gains what so ever for your data storage
(probably large files?).

with database workload the accesses are mostly random and small, in these
circumstances both the number of disks (in high load situations) and the
access speed both help preformance.

Significant preformance gains will only be under heavy loads, in this
situation it would help to increase the number of disks.

with single accesses the faster spinning disks will have a slight advantage
in access time, but you should consider upgrading your controller.

Any modern single channel controller is more preformant than the aging PL
you have (even in singel ended mode, if you would keep using your ST39173W)

With enough fast disks your controller pretty fast becomes the bottelneck
(the faster cpus like the i960RN/RS on modern controllers shave a few
millisecond of access time, probably as much as you would gain by using
faster disks) and that would also lessen the parity bottelneck when writing
raid 5 data. This way you would also see an significant increase in the
preformance of your data storage)

if you wouls悲 stay with mylex, you would probaly not even have to reinstall
everything, because the PL with 3.x firmware uses the same methode to arange
data on the disks as more modern ones (but you should ask the mylex suport
this question and do an Backusp anyway)

I personaly would buy an modern single channel (or multi channel, if you
plan to use more than 15 disks) LVD controler like the mylex acceleraid 352
or an icp GDTx523RZ/RN controllerand unse only LVD Disks on it.

Id use the mylex with the ST39173W for data storage or sell them together.

-Bodo


"Walter L. Preuninger II" <taz...@hotmail.com> schrieb im Newsbeitrag
news:rU8L7.14078$YD.13...@news2.aus1.giganews.com...

0 new messages