Re: maildir_cache_page_size

[Thomas Glanzmann <sithglan@xxxxxxxxxxxxxxxxxxxx>]

Hi Nicolas,

> How can I do this? db_dump.html does not help me.

I don't know. This is from Brian Feldman <green@xxxxxxxxxxx> however. I
did some tests and db4 does only support a pagesize up2 16kbyte. And it
got better with higher values. So I use 16kbyte.

I attached my initial benchmark. First line is cache build up second one
is cached.  This was my typical 90Mbyte 32thousand message maildir. DB4
only supports blocksizes till 16kbyte but gdbm doesn't perform that much
better on bigger blocksizes.

This was on my SunFIRE 280R with 8kbyte pagesize.

Sincerely,
        Thomas

--- Begin Message ---

• To: Brian Feldman <green@xxxxxxxxxxx>
• Subject: Re: [PATCH] disabling stat() in the maildir hcache
• From: Thomas Glanzmann <sithglan@xxxxxxxxxxxxxxxxxxxx>
• Date: Sun, 23 May 2004 09:32:21 +0200
• In-reply-to: <20040523070309.GE51125@xxxxxxxxxxxxxxxxxx>
• References: <20040521144105.GA18518@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> <20040521152208.GA24821@xxxxxxxxxxxxxxxxxx> <20040521163548.GA8496@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> <20040521171230.GA31443@xxxxxxxxxxxxxxxxxx> <20040521202531.GA11702@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> <20040521212346.GA32363@xxxxxxxxxxxxxxxxxx> <20040522173713.GA8168@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> <20040523051634.GA51125@xxxxxxxxxxxxxxxxxx> <20040523063528.GB2760@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> <20040523070309.GE51125@xxxxxxxxxxxxxxxxxx>
• User-agent: Mutt/1.5.6i

Hi Brian,

> Well, I'm just proposing what is readily available now: the output
> that mutt -v shows you.

The problem is that this string is save in puts("..."); so we can't use
it. :-(

I fixed a little glitch: The page size has to be short not a int. DB4
ignored the size, but gdbm started to create 400Mbyte big databases. :-)

here are my benchmarking results:

bs      gdbm                    db-4

1024    real    0m17.047s       real    0m21.656
        real    0m2.933s        real    0m3.140s

2048    real    0m16.499s       real    0m29.033
        real    0m2.935s        real    0m3.270s

4096    real    0m17.760s       real    0m21.380
        real    0m2.927s        real    0m2.826s

8192    real    0m18.656s       real    0m18.344
        real    0m2.620s        real    0m3.062s

16384   real    0m20.814s       real    0m17.710
        real    0m2.553s        real    0m3.631s


        Thomas

--- End Message ---