Subject: Re: tuning for small memory machines
To: Daniel Carosone <dan@geek.com.au>
From: Perry E. Metzger <perry@piermont.com>
List: tech-kern
Date: 12/02/2002 21:00:10
Daniel Carosone <dan@geek.com.au> writes:
> On Tue, Dec 03, 2002 at 10:53:45AM +1100, matthew green wrote:
> 
> > the "min" values stop pageout from happening, but if there is free
> > ram available any type of ram can use it up to it's "max", i believe.
> 
> That's how I understood it too. 
> 
> Max stops new pages getting allocated. Min stops more pages getting
> stolen, but if nobody wants pages of that type the memory can be
> used for other things.

In practice, this rarely happens. Remember, it isn't only active pages
that get cached this way. If you type "ls", ls's pages remain in even
if you're not actively using it.

> Perhaps that points to the problem - setting max too low results
> in pages going unused because noone wants them for the other types.

No, that really isn't an issue. I've experimented with this a lot and
the behavior nearly fully complies with one's theoretical expectation.

-- 
Perry E. Metzger		perry@piermont.com