From mboxrd@z Thu Jan  1 00:00:00 1970
From: ron minnich <rminnich@lanl.gov>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] Rstat needs three size fields?
In-Reply-To: <009d01c3454a$2b1ca5a0$d2944251@insultant.net>
Message-ID: <Pine.LNX.4.44.0307080812450.25149-100000@maxroach.lanl.gov>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Date: Tue,  8 Jul 2003 08:14:40 -0600
Topicbox-Message-UUID: ef0beeba-eacb-11e9-9e20-41e7f4b1d025

On Tue, 8 Jul 2003, boyd, rounin wrote:

> > Another common strategy that I've used is to double up until to
> > threshold, and then extend by a constant chunk size after that.  For
> > instance, you might double things until you get to, say, a megabyte,
> > and then add on a megabyte after that.
>
> yeah, that's not a bad idea, but it depends on what you're trying to
> allocate etc ...  defining the 'threshold' and the 'chunk size' is the
> hard part.

If the stuff is files, you can depend on some things. Every time I check a
file system I find the lovely bimodal size distribution. You can start
with 32K or some nominal such size and you'll cover most of the files;
past that point things are "big" so you can start allocating reasonably
big chunks.

That's how I've done it anyways, unless I pre-computed the size.

ron