On Tue, 11 Jul 2000, Andrea Arcangeli wrote:
> On Tue, 11 Jul 2000, Rik van Riel wrote:
> >On Tue, 11 Jul 2000, Andrea Arcangeli wrote:
> >> On Tue, 11 Jul 2000, Rik van Riel wrote:
> >>
> >> >No. You just wrote down the strongest argument in favour of one
> >> >unified queue for all types of memory usage.
> >>
> >> Do that and download an dozen of iso image with gigabit ethernet
> >> in background.
> >
> >You need to forget about LRU for a moment. The fact that
> >LRU is fundamentally broken doesn't mean that it has
> >anything whatsoever to do with whether we age all pages
> >fairly or whether we prefer some pages over other pages.
> >
> >If LRU is broken we need to fix that, a workaround like
> >your proposal doesn't fix anything in this case.
>
> So tell me how with your design can I avoid the kernel to unmap anything
> while running:
>
> cp /dev/zero .
>
> forever.
>
> Whatever aging algorithm you use if you wait enough time the
> mapped pages will be thrown away eventually.
And that is correct behaviour. The problem with LRU is that the
"eventually" is too short, but proper page aging is as close to
LFU (least _frequently_ used) as it is to LRU. In that case any
page which was used only once (or was only used a long time ago)
will be freed before a page which has been used more often
recently will be.
This effectively and efficiently protects things like X, xterm
and other things which are used over and over again, while still
swapping out things which are not used at all.
regards,
Rik
--
"What you're running that piece of shit Gnome?!?!"
-- Miguel de Icaza, UKUUG 2000
http://www.conectiva.com/ http://www.surriel.com/
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Jul 15 2000 - 21:00:13 EST