When the PC known as jogger-egg.com was new in late 1998, I measured
its raw disk read speed. It got about 12 Mb/sec doing large block
sequential reads from /dev/hda. This was about 30% faster than the
1996 vintage SCSI disk on my Octane at work. Cool. Exciting acronyms
like UDMA and ATA/33 were bandied about.
A few months later, I replaced jogger-egg's 14 Gb disk with a 21 Gb
disk and remeasured the speed. The new disk did 17 Mb/sec. Very
cool. At this time, I was still running SuSE 5.0 or 5.2.
Last week, I ran out of disk space again, and installed a whopping* 46
Gb disk. I happened to copy a 1 Gb file, and it took for*EVER*. I
reran the measurements that I'd done with the other disks. 4 Mb/sec.
Remeasured the 21 Gb disk. Also 4 Mb/sec. Uncool. Square. Dorky,
even.
Anyway, I found this interesting snippet in
/etc/rc.d/init.d/mandrake_everytime.
## Optimisation of Hard drive.
if grep -qi opti /proc/cmdline; then
if [ -x /sbin/hdparm ];then
echo hdparm
LIST_HD=$(grep '^hd.:' /var/log/dmesg|\
grep -ivE '(CD.*ROM|FLOPPY|TAPE|STATUS)'|cut -d:
-f1|sort|uniq)
if grep -i nohdparm /proc/cmdline >/dev/null ; then
action "Hard Drive optimisations disabled" \
echo ""
else
for i in $LIST_HD;do
action "Starting Hard Drive optimisations for $i" \
hdparm -q -c1 -q -A1 -q -m16 -q -d1 /dev/$i
done
fi
fi
fi
Sure enough, typing in that ugly hdparm command speeds the disks
right up. The new disk reads at 21.5 Mb/sec now.
I do not have "opti" in /proc/cmdline, so the hdparm command is never
executed. /proc/cmdline comes indirectly from /etc/lilo.conf. Does
anybody know exactly what I should have in /etc/lilo.conf to enable
IDE disk optimizations?
Thanks.
*see _Life,_the_Universe_and_Everything_ by Douglas Adams for
discussion of the verb, "to whop".
--
K<bob>
[EMAIL PROTECTED], http://www.jogger-egg.com/
