[29719] in North American Network Operators' Group
Re: LoadBalancing products: Foundry ServerIron
daemon@ATHENA.MIT.EDU (Bennett Todd)
Thu Jul 6 13:00:41 2000
Date: Thu, 6 Jul 2000 12:49:04 -0400
From: Bennett Todd <bet@rahul.net>
To: Karyn Ulriksen <kulriksen@publichost.com>
Cc: "'nanog@merit.edu'" <nanog@merit.edu>
Message-ID: <20000706124904.P486@oven.com>
Mime-Version: 1.0
Content-Type: multipart/signed; micalg=pgp-md5;
protocol="application/pgp-signature"; boundary="4IFtMBbmeqbTM/ox"
Content-Disposition: inline
In-Reply-To: <0127E258EE29D3118A0F00609765B448317873@subnet-gw-00053.sitestream.net>; from kulriksen@publichost.com on Thu, Jul 06, 2000 at 09:40:36AM -0700
Errors-To: owner-nanog-outgoing@merit.edu
--4IFtMBbmeqbTM/ox
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
So I think we're in complete agreement, even though we have opposite
things to say about the box.
For bandwidths up to say 50-60 Mbps they work beautifully. If you
want to go faster, you probably have to sacrifice some of the
smarts they bring to the job.
But while I certainly understand that circumstances will vary, from
my own experience I wouldn't try and scale a single load-balancer up
past say 50Mbps sustained capacity; I'd scale up from there by
deploying diverse server farms, plugged into substantially different
parts of the internet.
How many places are there where you can really usefully push more
than 50Mbps into one point of the internet without it just swelling
up into a big blister and popping? I realize that there are probably
plenty of private backbones that would have no trouble distributing
that sort o' bandwidth, but I've had bad luck with peering points;
whenever I see bad performance from one of my server farms it seems
like the real problem has ended up being where the provider's
backbone hooks up to other providers'.
Then again, that may just mean that I've been stuck using poorer
providers.
-Bennett
--4IFtMBbmeqbTM/ox
Content-Type: application/pgp-signature
Content-Disposition: inline
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.0 (GNU/Linux)
Comment: For info see http://www.gnupg.org
iD8DBQE5ZLiAL6KAps40sTYRAvZEAJ9XRstQ6UkW4Zgr28LJKRLeaMj6jACfRDEt
MmxivSCufY5AbUBdyQ094gY=
=uUYz
-----END PGP SIGNATURE-----
--4IFtMBbmeqbTM/ox--
