Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 7 Jul 2000 09:00:01 -0700 (PDT)
From:      Matthew Dillon <dillon@apollo.backplane.com>
To:        Chris Richards <crichard-freebsd@wso.williams.edu>
Cc:        freebsd-arch@FreeBSD.ORG
Subject:   Re: Alterations to vops
Message-ID:  <200007071600.JAA00696@apollo.backplane.com>
References:  <Pine.BSF.4.05.10007062327020.68909-100000@login-1.eunet.no> <200007070143.SAA96248@apollo.backplane.com> <20000706184859.C25571@fw.wintelcom.net> <20000706223150.A24949@klapaucius.zer0.org> <20000707080934.I25571@fw.wintelcom.net> <20000707114411.A31035@wso.williams.edu>

next in thread | previous in thread | raw e-mail | index | archive | help

:On Fri, Jul 07, 2000 at 08:09:34AM -0700, Alfred Perlstein wrote:
:
:> The only problem is that for the "average case" you waste nearly
:> 3.5k per directory cached, so it's only really a win on systems
:> with large amounts of ram.
:
:In your estimation, how large is "large"?  Of course each site should
:test to see what best suits its needs, but I'm looking for a rough
:idea.
:
:-chris

    The number is also not correct.  Sure, on a directory by directory
    basis the wastage can be 3.5K.  But if the page is cached in the 
    VM Page cache it will fall under the center-weighted LRU algorithm
    which means that idle directory cache pages will be reused by the 
    system fairly quickly.

    In contrast, (512 byte) directory blocks sitting the buffer cache's
    MALLOC space are not governed by standard cache reuse mechanisms and
    are not as reusable.

    My personal feeling is that the simple fact that the pages become 
    governed by standard reuse algorithms when in the VM Page cache more
    then compensates for the 'wasted' memory.

    And what does 'memory waste' mean anyway?  If the system performs
    better then we shouldn't really care that 3.5K out of 4K is wasted
    for a few seconds, verses 512 bytes 'not wasted' but also held in
    the malloc cache 'forever' (except when the malloc cache gets saturated,
    in which case you reuse the blocks but are screwed anyway because
    the malloc cache is too small).

					-Matt
					Matthew Dillon 
					<dillon@backplane.com>


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-arch" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200007071600.JAA00696>