CLOCK-Pro algorithm

CLOCK-Pro algorithm

[Adrian McMenamin]

I originally tried to send this to the addresses for Song Jiang, Feng
Chen and Xiaodong Zhang on the USENIX paper but it bounced from all of
them. So I hope you will indulge me if I send it to the list in the
hope it might reach them. Or perhaps someone here could answer the
questions below.

Many thanks

Adrian

Dear all,

I am just beginning work on an MSc project on Linux memory management
and have been reading your paper to the 2005 USENIX Annual Technical
Conference. I was wondering what the current status of this algorithm
is as regards the Linux kernel.

I can find this: http://linux-mm.org/ClockProApproximation and patches
for testing with the 2.6.12 kernel but am not entirely clear as to
whether this algorithm was included: certainly all the books I have
read still talk of the LRU lists that are similar to the 2Q model.

Could you enlighten me?

Many thanks in advance

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom policy in Canada: sign http://dissolvethecrtc.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: CLOCK-Pro algorithm

[Mel Gorman]

On Thu, Dec 30, 2010 at 11:24:00PM +0000, Adrian McMenamin wrote:
> I originally tried to send this to the addresses for Song Jiang, Feng
> Chen and Xiaodong Zhang on the USENIX paper but it bounced from all of
> them. So I hope you will indulge me if I send it to the list in the
> hope it might reach them. Or perhaps someone here could answer the
> questions below.
> 
> Many thanks
> 
> Adrian
> 
> Dear all,
> 
> I am just beginning work on an MSc project on Linux memory management
> and have been reading your paper to the 2005 USENIX Annual Technical
> Conference. I was wondering what the current status of this algorithm
> is as regards the Linux kernel.
> 
> I can find this: http://linux-mm.org/ClockProApproximation and patches
> for testing with the 2.6.12 kernel but am not entirely clear as to
> whether this algorithm was included: certainly all the books I have
> read still talk of the LRU lists that are similar to the 2Q model.
> 

The current reclaim algorithm is a mash of a number of different
algorithms with a number of modifications for catching corner cases and
various optimisations. In terms of an MSc, your best bet is to do a
general literature review of replacement algorithms and then do your
best to write a short paper describing the Linux page replacement
algorithm identifying which replacement algorithms it takes lessons
from.

-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom policy in Canada: sign http://dissolvethecrtc.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: CLOCK-Pro algorithm

[Adrian McMenamin]

[-- Attachment #1: Type: text/plain, Size: 1197 bytes --]

On 6 January 2011 10:46, Mel Gorman <mel@csn.ul.ie> wrote:

>
>
> The current reclaim algorithm is a mash of a number of different
> algorithms with a number of modifications for catching corner cases and
> various optimisations. In terms of an MSc, your best bet is to do a
> general literature review of replacement algorithms and then do your
> best to write a short paper describing the Linux page replacement
> algorithm identifying which replacement algorithms it takes lessons
> from.
>
>
Thanks for this - I am indeed reading through the various papers and other
literature (and I bought your book), though my aim with the MSc project is
slightly more abitious than maybe you are suggesting: I want to look at how
important some of the issues that are identified as common problems with
global clock and similar replacement algorithms (eg a slow response to
changes in locality) and to test whether there are some heuristics from
local replacement policies that might address them, at least in theory.

The Clock-Pro paper was an interesting read and given its claims for
improvement in the 2.4 series kernels I was interested in seeing how far the
idea had got in the 2.6 series.

Adrian

[-- Attachment #2: Type: text/html, Size: 1604 bytes --]