lkml.org 
[lkml]   [1997]   [Feb]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: Improving HardDisks...
Gabriele Paciucci <paciucci@pcainf1.ing.uniroma1.it> wrote:
> 1)Why don't Linux use the Bios settings ?
> 2)Why -d1 -X34 decrease ?
> 3)How to set -m16 -c3 -X12 to HD finally ?

A2. -d1 -X34 is supposed to put it into DMA mode 2, and using_dma should be 1
if it did. It's 0, so something went wrong I guess.
A3. Put it in your /etc/rc.d/rc.local

Note that throughput is not everything : some of those options don't
affect throughput much, but they reduce "system time" i.e. time spent in
kernel mode for interrupt handling etc. This is a big win since some other
process could run in the time which otherwise would be spent in disk I/O.

The following very rough measurement was done on a P-100 8MB IDE Seagate
ST31276A, kernel 2.0.29 with the Triton stuff, after booting with
init=/bin/bash. I haven't averaged or anything, but these are fairly
representative :

With default settings (everything off except readahead)
time hdparm -t /dev/hda

Throughput 3.66 MB/sec
Time 0.06user 3.55system 0:07.31elapsed 49%CPU
No. of ide0 interrupts 32810

With hdparm -c1 -m16 -d1 -X34 /dev/hda

Throughput 4.69 MB/sec
Time 0.04user 0.88system 0:06.59elapsed 13%CPU
No. of ide0 interrupts 260 !!

This is a HUGE order of magnitude improvement ! This deserves to be better
documented ... right now it's just in the hdparm manpage AFAIK. I don't think
too many people know about this ...

-- ganesh

\
 
 \ /
  Last update: 2005-03-22 13:38    [W:0.021 / U:0.692 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site