If just copying a partition, then you can pretty much ignore ATA if you are
above 66 -
the thing that matters is :
fragmentation, RPM, and physical sectors per track, followed by tracks per
cylinder
63 sectors of 512 bytes per track = 30K per revolution
If RPM is 7200, that's 120 rps = 120x30K of data per second
( ignoring head movement time, on the basis that the drive is not
fragmented, and you are accessing the files in the sequence that they were
written onto the drive)
and you were doing an image anyhow
So the physically moving part of the drive can manage UPTO 3.6Mb/second
( That's the maximum sustained throughput - = 30Mbits per second)
Physical read time is at least drive size/max sustained read speed
do your own maths but remember to use the correct number of sectors
( divide drive capacity by cylinders and used platters - you probably won't
find the noumber is anywhere as low as 63)
Next consideration -
How many drives are (trying to ) using the interface cable at the same time
lose 25% if two drives are actively using the cable at the same time
They need to negociate which can use it, and then get confirmation that it's
them
Now - are they using DMA or PIO
PIO uses the CPU to move data from the drive controller to memory
DMA doesn't
So that can add delays to the transfer of data to memory,
any small, but regulear delay can cause no read action to happen during each
alternate revolution, the first sector of a track having just passed the
heads when the read request is processed by the drive controller
and we cannot even begin to consider the delays if the heads have to go
somewhere else for allocation table, directory, or just the next portion of
a file
Now - you have to get the data down the USB 1 cable - into the drive,
USB 1.1 transfer rate
add in directory creation and access for each file, as well as updating the
allocation table
Using delayed write? to avoid each change being written to the drive before
the next action
Not using delayed write because the USB connection is not as reliable as
Firewire S-ATA, or plain P-ATA
Ballpark figures, - no write delay
Expect the write process to use about 4 extra revolutions for each file
written, plus another 2 for each allocation unit for the file
(head movements allowed for during the drive revolution.)
Yes - you were doing an image anyhow..
BUT did you have the drive verified - condition prior to use, and data
checked after use, was it formatted before the image copy
If you can, consider an additional PCI board to run S-ATA with hot-swap for
drives in a caddy
that way you get 150 transfer rate between the controller and memory
and - if you are lucky, you will find that transfers to/from the drives can
be concurrent, rather than alternately as is usually the case with the
integrated IDE controllers
P.S. what is the max transfer rate of USB 1.1? 1.5 Mbytes per sec
USB 2 is 60Mbytes per sec MAXIMUM
see
PCWorld.com - News & Trends: USB 2.0's Real Deal
JimB
--
----------------------------------------
WIN-HOME Archives: http://PEACH.EASE.LSOFT.COM/archives/WIN-HOME.html
Contact the List Owner about anything: [EMAIL PROTECTED]
Official Win-Home List Members Profiles Page
http://www.besteffort.com/winhome/Profiles.html
