On Nov,Tuesday 13 2007, at 1:18 AM, Steven M. Bellovin wrote:

> On Mon, 12 Nov 2007 20:38:10 +0000 (UTC)
> mlelstv@serpens.de (Michael van Elst) wrote:
>
>> jmarin@embedtronics.fi (Jukka Marin) writes:
>>
>>> Wow :-)  But I don't think delaying a catastrophe is the same as
>>> preventing it :-)  I would like to slow down the sync operation (to
>>> leave part of the I/O capacity for other things).  I guess I could
>>> try making the delays real short and see what happens then..
>>
>> The problem comes from running out of buffers. You want to sync
>> often and you want to sync fewer buffers in a burst to mitigate
>> the problem.
>
> This, I think, is a large part of the actual solution to the problem.
> The other part is to have the process scheduler penalize processes  
> -- or
> users? -- that do too much I/O.


I thought that this years SOC projects was QOS inside UVM [1].
This project should implement QOS which penalise writes against readers
do what you want for processes. In this years SOC report I read that
Summatra Kundu was ill in summer but after the summer he started work
on this project again. I don't know in what state is this project now.

http://netbsd-soc.sourceforge.net/projects/qos/

Regards

Adam.