Date: Wed, 20 Mar 2002 20:01:04 +0000 From: David Malone <dwmalone@maths.tcd.ie> To: Mike Silbersack <silby@silby.com> Cc: cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: src/sys/ufs/ufs ufs_dirhash.c Message-ID: <200203202001.aa48408@salmon.maths.tcd.ie> In-Reply-To: Your message of "Wed, 20 Mar 2002 12:28:38 CST." <20020320122802.U54496-100000@patrocles.silby.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> > mean #probes for
> > home dir mh inbox news/cancel tmp squid2 squid3
> > old successful 1.02 3.19 4.07 1.10 7.85 2.06
> > new successful 1.04 1.32 1.27 1.04 1.93 1.17
> >
> > old unsuccessful 1.08 4.50 5.37 1.17 10.76 2.69
> > new unsuccessful 1.08 1.73 1.64 1.17 2.89 1.37
> What do the numbers look like w/o DIRHASH? I'm curious...
The closest thing I could produce to a direct comparison would be
the number of blocks fetched from the buffer cache. For a successful
lookup this is (dirsize/2) and for an unsuccessful lookup (dirsize).
For the table above:
home dir mh inbox news/cancel tmp squid2 squid3
unsuccessful 47 58 3987 24 8 13
successful 23.5 29 1993.5 12 4 6.5
This comparison isn't exact 'cos dirhash won't always have to fetch
a block for each probe and it also knows where to look within the
block (without dirhash you always start at the beginning of the
block). This also doesn't factor in the initial building of the
hash or the nami cache.
If you access the entries of the directory in the order readdir
returns them, then the number of probes will be 1 in both the dirhash
and non-dirhash case. This is what most sensible programs do.
David.
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi? <200203202001.aa48408>