Hard disk sustained throughput

  • Thread starter Thread starter PeteOlcott
  • Start date Start date
P

PeteOlcott

How fast should a SATA hard-drive be able to copy a 16 GB folder from
one disk to another?
http://www.nextlevelhardware.com/storage/barracuda/

This link (Folder Copy Test) indicates that 5 minutes is about the
maximum speed when there are few
large files. My machine took ten times this long on 12.3 GB of data in
196,916 files in 48,761 folders. Is this a reasonable amount of time
to copy data from one SATA drive to another?
 
PeteOlcott said:
How fast should a SATA hard-drive be able to copy a 16 GB folder
from one disk to another?

http://www.nextlevelhardware.com/storage/barracuda/

This link (Folder Copy Test) indicates that 5 minutes is about
the maximum speed when there are few large files. My machine took
ten times this long on 12.3 GB of data in 196,916 files in 48,761
folders. Is this a reasonable amount of time to copy data from one
SATA drive to another?

Doesn't sound unreasonable. Consider that the system had to create
those files, create and expand the directories involved, find the
storage, etc. That was still about 3900 files per minute, ignoring
directory creation.
 
PeteOlcott said:
How fast should a SATA hard-drive be able to copy a 16 GB folder from
one disk to another?
http://www.nextlevelhardware.com/storage/barracuda/

This link (Folder Copy Test) indicates that 5 minutes is about the
maximum speed when there are few
large files. My machine took ten times this long on 12.3 GB of data in
196,916 files in 48,761 folders. Is this a reasonable amount of time
to copy data from one SATA drive to another?

On Vista or XP?
On Vista - get the service pack!!
 
PeteOlcott said:
This is on XP x64

A folder copy test will have a strong "seek" component to the copy
time. The sustained transfer rate of a drive is not the only consideration.
If you had 196,916 2KB files, the transfer time when the head gets there,
would be virtually zero for each file, but it takes milliseconds to find each
file ("seek" the heads to the appropriate track). If you can process
500 of the 2KB files each second, due to the limitations of head movement,
that is 1MB/sec transfer rate, and looks pathetic. And this is where a flash
drive would excel, as the seek time is very low by comparison. (USB flash
is 1 millisecond seek, while IDE or SATA based flash can be 1/10th of that.)

Paul
 
Paul said:
A folder copy test will have a strong "seek" component to the copy
time. The sustained transfer rate of a drive is not the only consideration.
If you had 196,916 2KB files, the transfer time when the head gets there,
would be virtually zero for each file, but it takes milliseconds to find
each
file ("seek" the heads to the appropriate track). If you can process
500 of the 2KB files each second, due to the limitations of head movement,
that is 1MB/sec transfer rate, and looks pathetic. And this is where a
flash
drive would excel, as the seek time is very low by comparison. (USB flash
is 1 millisecond seek, while IDE or SATA based flash can be 1/10th of
that.)

Paul
http://www.geocities.com/vgrinenko/DiskSpeed32/
You can use something like this disk speed test to get an general idea
of the speed of a drive. Still Paul has a point, seek speed, buffer
size, defrag status, etc all factor into the overall speed.
 
Paul said:
A folder copy test will have a strong "seek" component to
the copy
time. The sustained transfer rate of a drive is not the
only consideration.
If you had 196,916 2KB files, the transfer time when the
head gets there,
would be virtually zero for each file, but it takes
milliseconds to find each
file ("seek" the heads to the appropriate track). If you
can process
500 of the 2KB files each second, due to the limitations
of head movement,
that is 1MB/sec transfer rate, and looks pathetic. And
this is where a flash
drive would excel, as the seek time is very low by
comparison. (USB flash
is 1 millisecond seek, while IDE or SATA based flash can
be 1/10th of that.)

Paul

Yes that makes complete sense, so it is quite reasonable
that copying folders of 200,000 small files takes
fifteen-fold longer than copying the same amount of data
comprised of 100 very large files.

Do you know why deleting many small files could take a much
more unreasonable amount of time on Windows XP x64?
At the rate the deletion was going it would have taken 20
hours to delete about 200,000 files. I ended up reformatting
the drive and reloading the data, this took much less time.
 
Peter said:
Yes that makes complete sense, so it is quite reasonable
that copying folders of 200,000 small files takes
fifteen-fold longer than copying the same amount of data
comprised of 100 very large files.

Do you know why deleting many small files could take a much
more unreasonable amount of time on Windows XP x64?
At the rate the deletion was going it would have taken 20
hours to delete about 200,000 files. I ended up reformatting
the drive and reloading the data, this took much less time.

I haven't a clue. I don't know if there is any way, short of
using some kind of development environment debugger, to
find out what is going on.

Some problems are obvious ones. There was an SIS chipset and
driver, where when you clicked on an icon, there was no
response from the computer for five seconds, and then the
operation would start. In that case, it seemed the
first file operation would fail (no response), time out,
and get retried. At least that has obvious visible symptoms.
(Nice consistent 5 second delay, no CPU utilization to speak of
until the five seconds is up.) But for more subtle cases
(like yours - only being able to delete 3 files per second),
it could be something hidden, like antivirus software activity.

Sysinternals.com has some tools for watching what is going on,
but if there were issues at a driver level, I don't know if
you'd see anything incriminating or not. I cannot play with
just any Sysinternals tool on my computer here, because my
antivirus software (Kaspersky) gets into a fight with some of
the programs, resulting in a computer freeze that not even
control-alt-delete can fix. (Control-alt-delete is "hooked" by
the AV software, and if the AV is "busy", then control-alt-delete
stops working. Talk about invasive.)

Paul
 
Peter said:
Yes that makes complete sense, so it is quite reasonable
that copying folders of 200,000 small files takes
fifteen-fold longer than copying the same amount of data
comprised of 100 very large files.

Do you know why deleting many small files could take a much
more unreasonable amount of time on Windows XP x64?
At the rate the deletion was going it would have taken 20
hours to delete about 200,000 files. I ended up reformatting
the drive and reloading the data, this took much less time.

I figured I'd pass this along, as this was referenced in another
group tonight.

"Increase XP NTFS performance - NtfsDisableLastAccessUpdate, fiddle with MFT"
http://www.tweakxp.com/article37043.aspx

Paul
 
Back
Top