> First approach on my laptop, quite significant.
> Comparison of r/w from/to :
>  - the utility partition on the hd (fat32)
> - the usb key
> shows that read from the usb key is about 2 times slower than from the hard
> disk, which is expected, whereas the write is about 20 times slower, which 
> seems
> in deed very wrong.

> # ls -l Desktop/R174292/Dell_multi-device_A17_R174292.exe
> -rw-r--r-- 1 root root 56396160 Mar 29 15:48 
> Desktop/R174292/Dell_multi-device_A17_R174292.exe

> copy to usb 2.0 key fat32
> # time cp Desktop/R174292/Dell_multi-device_A17_R174292.exe /media/z-PEN/
> real  1m58.760s
> user  0m0.009s
> sys   0m0.358s

Hmm, so it transferred less than 500 kbytes/second,
for that 56mbyte file copy.

Somehow your write performance of 500 kbytes/second
reminds me on an interesting usb flash memory performance
issue I see with one of my usb flash memory devices:

Apr  7 16:36:08 tiger2 usba: [ID 912658 kern.notice] USB 2.0 device 
(usb1976,6025) operating at hi speed (USB 2.x) on USB 2.0 root hub: [EMAIL 
PROTECTED], scsa2usb1 at bus address 3
Apr  7 16:36:08 tiger2 usba: [ID 349649 kern.notice]    Alliance Flash Disk 
17164900F0B50800

The format of my usb stick's device identification looks similar to 
the one you reported - both devices seem to use the same
16 character hex serial number format.  In prtconf -Dv output
I found these for my usb flash memory device:

usb-vendor-name: 'Alliance'
usb-product-name: 'Flash Disk'
usb-serialno: '17164900F0B50800'
usb-vendor-id: 00001976
usb-product-id: 00006025
usb-revision-id: 00000100


I guess your device uses vendor-name = 'DaneElec',
product-name = 'z-PEN', and
serialno = '0DF1A96051D22BC1'


Now the interesting thing with my 'Alliance' 'Flash Disk'
is that it is able to write reasonable fast to the raw device
using wrte block sizes of 1K, 2K and 8K (2.3 mbytes/sec), but
write transfer rates gets really bad when using write block
sizes of 4K!  But with 4K writes transfer rates to the raw
device drop to 500 kbytes/sec.

Like this, when writing 100mbytes to the raw usb flash memory device:

# ptime dd if=flash4g.img of=/dev/rdsk/c6t0d0p0 bs=8k count=12800
12800+0 records in
12800+0 records out

real       45.095
user        0.041
sys         0.641

# ptime dd if=flash4g.img of=/dev/rdsk/c6t0d0p0 bs=4k count=25600
25600+0 records in
25600+0 records out

real     3:30.229
user        0.041
sys         0.521

# ptime dd if=flash4g.img of=/dev/rdsk/c6t0d0p0 bs=2k count=51200
51200+0 records in
51200+0 records out

real       46.831
user        0.046
sys         0.569

# ptime dd if=flash4g.img of=/dev/rdsk/c6t0d0p0 bs=1k count=102400
102400+0 records in
102400+0 records out

real       47.742
user        0.082
sys         0.975


I do remember that the preformatted fat filesystem on my usb flash
memory stick used a 4K cluster size, and that writing to that stick
using pcfs was extremely slow at first.  As a workaround I reformatted
the stick to use a cluster size of 8K, and IIRC I also tried to add the
correct amount of hidden sectors so that the 8K cluster are aligned
on 8K boundaries on the physical disk media.
 
 
This message posted from opensolaris.org
_______________________________________________
opensolaris-discuss mailing list
[email protected]

Reply via email to