So I got bored enough to do some testing with external drives.  SCSI is
a bit expensive, USB is slow, and Firewire is still in progress.

Firewire is the best choice of [USB, USB2, Firewire] as FW has a top
speed of 400MB/s, and drivers are included with the more recent 2.4
releases.  USB tops out at 12Mb, USB2 is 480MB, but the drivers are only
available in 2.5.x kernels.  Also, availability of USB2 cards is more
limited than FW, since you can usually find FW items in the
Mac section of stores or catalogs.

I picked up a Siig FW card (PCI, 3 external ports) for $39, and an ACOM
60GB external drive for $229.  Like the external USB drives, these
drives are not self-powered, meaning you have to have a spare power plug
available.  Fortunately, the power supply is built into the drive, so
no bulky transformers.  You can probably pick up the chassis alone
and add your own IDE HDD or CD-R for about $99.

Both USB and FW drives show up as SCSI drives, meaning you have extra
overhead of going through the FW/USB drivers, then through the SCSI
drivers, and so on.

So here's some of the testing I have.  System is a Celeron 700 running
Debian with the 2.4.17 kernel.  Latest IEEE1394 drivers from
http://linux1394.sourceforge.net 

The first number of the result of hdparm -Tt is the timing of the cache
to memory which doesn't appear to have much difference between the
drives.  The second number is what you actually want to pay attention
to.

IDE drive:
# hdparm /dev/hda

/dev/hda:
 multcount    = 16 (on)
 I/O support  =  3 (32-bit w/sync)
 unmaskirq    =  0 (off)
 using_dma    =  1 (on)
 keepsettings =  0 (off)
 nowerr       =  0 (off)
 readonly     =  0 (off)
 readahead    =  8 (on)
 geometry     = 1216/255/63, sectors = 19541088, start = 0
 busstate     =  1 (on)

# hdparm -Tt /dev/hda

/dev/hda:
 Timing buffer-cache reads:   128 MB in  1.10 seconds =116.36 MB/sec
 Timing buffered disk reads:  64 MB in  3.56 seconds = 17.98 MB/sec

Pretty impressive.  Actually pretty good as I've got DMA turned on.
With DMA turned off, I was getting about 3MB/s.  Using multicount
and I/O support does not have much effect, but I have them turned
on anyway.

Now for the USB drive.  Just a plain 'ol USB chassis with a 30GB WD
drive in.

# hdparm -Tt /dev/sda

/dev/sda:
 Timing buffer-cache reads:   128 MB in  1.18 seconds =108.47 MB/sec
 Timing buffered disk reads:  64 MB in 86.87 seconds =754.41 kB/sec

Yow!  Real slow.  On the good side, I'm just storing MP3s on here.  Here
is also the reason why most USB 1.x CD-ROMs are limited to 4x
(600kB/sec).

Now for the Firewire drive.  400MB, here we come!

# hdparm -Tt /dev/sdb

/dev/sdb:
 Timing buffer-cache reads:   128 MB in  1.11 seconds =115.32 MB/sec
 Timing buffered disk reads:  64 MB in  9.56 seconds =  6.69 MB/sec

Hey!  This is half as slow as IDE!  What gives?  Well, part of it is
that I needed the latest drivers to work with the ACOM drive, otherwise
the SCSI drivers think the drive is a scanner or printer.  I also think
some of the overhead of FW/SCSI drivers is another part.

The thing to remember about this, though, is you can daisy-chain FW
drives all you want.  The external chassis I have has two FW ports on it
so you can use multiple drives on the same FW chain, and using multiple
drives at once will not cause the troubles that multiple drives on a IDE
chain will.

I would add some dd timings, but I'm using a mix of EXT3 and EXT2, which
would taint the results somewhat.  When I got all EXT3 (probably next
time I get around to rebooting) I may try it.

In the end, FW is a pretty nice way to expand the storage of your system
without requiring a reboot (if you already have the drivers), not have
to pay for SCSI prices, and not fool around with IDE drives.  The cost
to entry is pretty low (cards run from $39-$99, 60G drive for $229) and
the performance is adequate and hopefully improving.

-Mark


*****************************************************************
To unsubscribe from this list, send mail to [EMAIL PROTECTED]
with the text 'unsubscribe gnhlug' in the message body.
*****************************************************************

Reply via email to