Re: kernel 2.4.21 , large disk write => system crawls

[Jakob Oestergaard <jakob@unthought.net>]

On Thu, Jul 03, 2003 at 06:56:24PM -0400, Mark Hahn wrote:
> > >>when i do a large disk write operation ( copy a big file for example ), 
> > >>the whole system becomes very busy ( system goes into 99% cpu 
> 
> it's not write-specific.  you can see below that you're somehow
> managing to trigger roughly two interrupts per *either* bi or bo.
> for a normal IDE setup, you should see one interrupt per 16-64K
> under average use.  it's almost like your sys somehow thinks
> that it can only transfer 1 sector per interrupt!

The system is 90% in kernel when writing.  That's bad (unless you're
maxing out your PCI busses or main memory bandwidth of course ;)

The IDE disks are most likely not running DMA.

> 
> > everytime i experience a slowdown, there's a 'big' number in the io (bo) 
> > column.
> 
> no, it's basically in=2*(bi+bo), as if your system somehow believes
> it can only do a single sector per interrupt (PIO and -m1 perhaps?)
> it should be more like 32K per interrupt.

Try:

hdparm -d1 /dev/hda


-- 
................................................................
:   jakob@unthought.net   : And I see the elder races,         :
:.........................: putrid forms of man                :
:   Jakob Østergaard      : See him rise and claim the earth,  :
:        OZ9ABN           : his downfall is at hand.           :
:.........................:............{Konkhra}...............: