xfs
[Top] [All Lists]

Re: PARTIAL TAKE - shrink xfs inode cache size

To: Federico Sevilla III <jijo@xxxxxxxxxxx>
Subject: Re: PARTIAL TAKE - shrink xfs inode cache size
From: Steve Lord <lord@xxxxxxx>
Date: 04 May 2003 08:03:14 -0500
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: <20030504042006.GC31338@leathercollection.ph>
References: <200305021117.h42BHrt03492@jen.americas.sgi.com> <20030504042006.GC31338@leathercollection.ph>
Sender: linux-xfs-bounce@xxxxxxxxxxx
On Sat, 2003-05-03 at 23:20, Federico Sevilla III wrote:
> On Fri, May 02, 2003 at 06:17:53AM -0500, Steve Lord wrote:
> > XFS does not use the union at the end of the inode for its fs private
> > data. There is 160 bytes per inode here which is wasted. Since we
> > manage our own inode allocations, we can do something about this and
> > actually not allocate that space. Saved 10Mbytes of space on my
> > machine this morning after nightly cron jobs had pulled inodes into
> > memory.
> 
> I am curious: what did this cron job of yours do to pull inodes into
> memory, allowing the changes in XFS's inode handling to save disk space?
> Just so we know what to do, perhaps when we jump to 2.4.21 or something
> like that. :)

updatedb runs everynight and does a tree walk through the filesystems,
once inodes are in memory, then depending on your system, it can be
hard to get the memory back again. All I did was remove some space 
in each inode which xfs does not need, hence fitting more per page
and making them use less memory.

The 10 Mbytes was just the difference in memory consumption by inodes
between the old and new sizes.

Steve



<Prev in Thread] Current Thread [Next in Thread>