[11636] in www-talk@info.cern.ch
Re: Effectiveness of Proxy Caching today
daemon@ATHENA.MIT.EDU (Martin Hamilton)
Sun Jun 11 13:21:35 2000
To: "John Hardy" <jh@lagado.com>
Cc: www-talk@w3.org
In-Reply-To: Message from "John Hardy" <jh@lagado.com>
of "Fri, 09 Jun 2000 18:19:50 +1000." <004501bfd1eb$7d78ed20$fe482bcb@lagado.com>
Date: Sun, 11 Jun 2000 18:07:57 +0100
From: Martin Hamilton <martin@net.lut.ac.uk>
Message-Id: <E131BDJ-0005mf-00@gadget.lut.ac.uk>
Resent-From: www-talk@w3.org
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Content-Type: text/plain; charset=us-ascii
"John Hardy" writes:
| Does it follow that most page accesses are uncacheable? Is this compensated
| by static content such as images, MP3's etc
Hi John - it seems to be the case there's a substantial enough volume
of cacheable (but not necessarily static :-) content out there that
(at least one level of) proxy caching is still useful. For instance,
the first level (site) caches I help to run are shipping several
million URLs/day between them, and have been achieving byte and
request hit rates in the region of 45% to 55% of the requested URLs on
a combined cache (pooled using cache digests) of just 100GB.
They've been getting another ~15% hit rate from peerings with a second
level "national" cache cluster that (by way of a comparison) has some
1.5TB of pooled disk and ships ~80m URLs/day and ~800GB/day peak. More
at <URL:http://wwwcache.ja.net/events/JISC-2000/> in case anyone's
interested...
You might also want to check out the proceedings of the web caching
workshops - best to start at <URL:http://www.terena.nl/conf/wcw/>,
which has only just taken place and so has the freshest results :-)
Cheers,
Martin
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.1 (GNU/Linux)
Comment: Exmh version 2.1.1 10/15/1999 + martin
iD8DBQE5Q8dtVw+hz3xBJfQRAv74AJwLEMTVp2Y0YcHuMLWcKIcom8DBcwCdHYOz
pLzIk6Dj8SGvUaiust2zEXY=
=UTP5
-----END PGP SIGNATURE-----
