Re: Some thoughts about cache and swap

[Matt Mackall <mpm@selenic.com>]

On Sun, Jun 06, 2004 at 10:38:25AM +0200, Christian Borntraeger wrote:
> John Bradford wrote:
> > Quote from Rik van Riel <riel@redhat.com>:
> > > I wonder if we should just bite the bullet and implement
> > > LIRS, ARC or CART for Linux.  These replacement algorithms
> > > should pretty much detect by themselves which pages are
> > > being used again (within a reasonable time) and which pages
> > > aren't.
> > Is there really much performance to be gained from tuning the 'limited'
> > cache space, or will it just hurt as many or more systems than it helps?
> 
> Thats a very good question. 
> Most of the time the current algorithm works quite well.
> On the other hand, I definitely know what people mean when they complain 
> about cachingand all this stuff. By just copying a big file that I dont use 
> afterwards or watching an video I have 2 wonderful scenarios.

Perhaps people should read about the referenced algorithms. LRU
(including the hybrid LRU that Linux uses) is vulnerable to
"scanning" of the sort you're describing, while the above algorithms
have varying degrees of scan-resistance. As lack of scan-resistance
seems to be "the big problem" in the current VM, this looks like an
interesting direction to go in.

-- 
Mathematics is the supreme nostalgia of our time.