SATA 2 Native Command Cueing and Multi Tasking. Alot better ???

SATA 2 Native Command Cueing and Multi Tasking. Alot better ???

I know the speed of SATA 2 is 300mbs. Not to fussed about that as we
are not even using the ATA 100 full bandwitdth yet

But i am interested on what effects NCQ has on responsivness of your
computer system. Will it make it much smoother.

Example coying a file and doinf somthing. That make my system grind
toa halt. Still can use it but much easier to leave it and wait.

Iam waiting to build a new 64 bit system. An Amd X2 with SATA 2 driver
sounds IDEAL.

I never want to see an Hour glass or my system slow down like it does
now.

Thanks for any info.

As it has been for some time, the amount of CPU time required for IDE data
transfer is very small for a drive or two in use at the same time. Those
small slices begin to add up in a server having many active drives.

Your system doesn't grind to a halt because of CPU usage for hard drive
transfers, but there could be major slow downs because all applications
using the hard drive would incur delays because rotational delays and head
repositioning delays to access files at different locations on the drive.
Nothing will help that on a single drive system. If you have many hard
drives and many threads using hard drive I/O (file server, transaction
server), then response time will be improved by something lik NCQ

Phil Weldon

"Son Of Sheep." sheep.com.au wrote in message
...


SATA 2 Native Command Cueing and Multi Tasking. Alot better ???

I know the speed of SATA 2 is 300mbs. Not to fussed about that as we
are not even using the ATA 100 full bandwitdth yet

But i am interested on what effects NCQ has on responsivness of your
computer system. Will it make it much smoother.

Example coying a file and doinf somthing. That make my system grind
toa halt. Still can use it but much easier to leave it and wait.

Iam waiting to build a new 64 bit system. An Amd X2 with SATA 2 driver
sounds IDEAL.

I never want to see an Hour glass or my system slow down like it does
now.

Thanks for any info.

Phil Weldon wrote:

As it has been for some time, the amount of CPU time required for IDE data
transfer is very small for a drive or two in use at the same time. Those
small slices begin to add up in a server having many active drives.

Your system doesn't grind to a halt because of CPU usage for hard drive
transfers, but there could be major slow downs because all applications
using the hard drive would incur delays because rotational delays and head
repositioning delays to access files at different locations on the drive.
Nothing will help that on a single drive system.

I agree it's much more useful on server applications but there *is* a way
to help on a single drive: command reordering. Instead of taking commands
in time sequence, which might move the head from one end of the drive to
the other and then back to the middle, you reorder according to seek
positioning so it picks up the third request for a middle sector 'on the
way' to the other end of the drive for the second one.


If you have many hard
drives and many threads using hard drive I/O (file server, transaction
server), then response time will be improved by something lik NCQ

Phil Weldon

"Son Of Sheep." sheep.com.au wrote in message
...


SATA 2 Native Command Cueing and Multi Tasking. Alot better ???

I know the speed of SATA 2 is 300mbs. Not to fussed about that as we
are not even using the ATA 100 full bandwitdth yet

But i am interested on what effects NCQ has on responsivness of your
computer system. Will it make it much smoother.

Example coying a file and doinf somthing. That make my system grind
toa halt. Still can use it but much easier to leave it and wait.

Iam waiting to build a new 64 bit system. An Amd X2 with SATA 2 driver
sounds IDEAL.

I never want to see an Hour glass or my system slow down like it does
now.

Thanks for any info.

'David Maynard' wrote, in part:
| there *is* a way
| to help on a single drive: command reordering. Instead of taking commands
| in time sequence, which might move the head from one end of the drive to
| the other and then back to the middle, you reorder according to seek
| positioning so it picks up the third request for a middle sector 'on the
| way' to the other end of the drive for the second one.

That would certainly be useful, but what algorithm is used by the drives? A
transaction server application could should a big improvement with a simple
algorithm, but for a single drive and multiple applications, who ends up on
the short end of the stick? Perhaps some future desktop operating system
might help make use of such drives, but for now?

Phil Weldon


"David Maynard" wrote in message
...
Phil Weldon wrote:

As it has been for some time, the amount of CPU time required for IDE
data transfer is very small for a drive or two in use at the same time.
Those small slices begin to add up in a server having many active drives.

Your system doesn't grind to a halt because of CPU usage for hard drive
transfers, but there could be major slow downs because all applications
using the hard drive would incur delays because rotational delays and
head repositioning delays to access files at different locations on the
drive. Nothing will help that on a single drive system.

I agree it's much more useful on server applications but there *is* a way
to help on a single drive: command reordering. Instead of taking commands
in time sequence, which might move the head from one end of the drive to
the other and then back to the middle, you reorder according to seek
positioning so it picks up the third request for a middle sector 'on the
way' to the other end of the drive for the second one.


If you have many hard drives and many threads using hard drive I/O (file
server, transaction server), then response time will be improved by
something lik NCQ

Phil Weldon

"Son Of Sheep." sheep.com.au wrote in message
...


SATA 2 Native Command Cueing and Multi Tasking. Alot better ???

I know the speed of SATA 2 is 300mbs. Not to fussed about that as we
are not even using the ATA 100 full bandwitdth yet

But i am interested on what effects NCQ has on responsivness of your
computer system. Will it make it much smoother.

Example coying a file and doinf somthing. That make my system grind
toa halt. Still can use it but much easier to leave it and wait.

Iam waiting to build a new 64 bit system. An Amd X2 with SATA 2 driver
sounds IDEAL.

I never want to see an Hour glass or my system slow down like it does
now.

Thanks for any info.

Phil Weldon wrote:
'David Maynard' wrote, in part:
there *is* a way
to help on a single drive: command reordering. Instead of taking
commands in time sequence, which might move the head from one end of
the drive to the other and then back to the middle, you reorder
according to seek positioning so it picks up the third request for a
middle sector 'on the way' to the other end of the drive for the
second one.

That would certainly be useful, but what algorithm is used by the
drives? A transaction server application could should a big
improvement with a simple algorithm, but for a single drive and
multiple applications, who ends up on the short end of the stick?

This is one thing that differentiates SCSI drives from SATA drives, even the
Raptor. SATA drives are firmware-tweaked towards desktop/workstation use
(low concurrency), whereas SCSI drives are firmware-tweaked towards server
yse (lots of concurrent accesses). Some SCSI drives have an option of
switching between desktop and server mode, and this makes quite a change in
both server and workstation benchmarks (in the obvious direction). That's
why a Raptor is able to keep up with or beat a Cheetah 15K.3 (or even a
15K.4 if you run it in server mode) in workstation benchmarks despite having
significantly inferior raw specs, and also why it can't keep up with the
comparitively ancient (3 years older) Cheetah 73LP in server benchmarks
despite having better raw specs.

However, there's the typical chicken and the egg problem. Hardware
manufacturers don't want to spend time writing server firmware for a SATA
drive unless there's a market for such usage of SATA drives, and there's a
limited server market for SATA drives because they don't have a
server-tweaked firmware. However, with the appearance of SATA hot-pluggable
RAID cabinets, I'm sure this will improve in the future.

[...]

--
Michael Brown
www.emboss.co.nz : OOS/RSI software and more
Add michael@ to emboss.co.nz ---+--- My inbox is always open

Phil Weldon wrote:

'David Maynard' wrote, in part:
| there *is* a way
| to help on a single drive: command reordering. Instead of taking commands
| in time sequence, which might move the head from one end of the drive to
| the other and then back to the middle, you reorder according to seek
| positioning so it picks up the third request for a middle sector 'on the
| way' to the other end of the drive for the second one.

That would certainly be useful, but what algorithm is used by the drives? A
transaction server application could should a big improvement with a simple
algorithm, but for a single drive and multiple applications, who ends up on
the short end of the stick? Perhaps some future desktop operating system
might help make use of such drives, but for now?

I don't know what you mean by "who ends up on the short end of the stick?"
A sector request is a sector request. Doesn't matter what app made it.

Besides, I wasn't designing one. I was just dealing with the matter of
whether 'nothing' could be done with one drive.


Phil Weldon


"David Maynard" wrote in message
...

Phil Weldon wrote:


As it has been for some time, the amount of CPU time required for IDE
data transfer is very small for a drive or two in use at the same time.
Those small slices begin to add up in a server having many active drives.

Your system doesn't grind to a halt because of CPU usage for hard drive
transfers, but there could be major slow downs because all applications
using the hard drive would incur delays because rotational delays and
head repositioning delays to access files at different locations on the
drive. Nothing will help that on a single drive system.

I agree it's much more useful on server applications but there *is* a way
to help on a single drive: command reordering. Instead of taking commands
in time sequence, which might move the head from one end of the drive to
the other and then back to the middle, you reorder according to seek
positioning so it picks up the third request for a middle sector 'on the
way' to the other end of the drive for the second one.



If you have many hard drives and many threads using hard drive I/O (file
server, transaction server), then response time will be improved by
something lik NCQ

Phil Weldon