[63100] in North American Network Operators' Group
Re: Average case performance vs. Worst-case guarantee
daemon@ATHENA.MIT.EDU (Mark Rogaski)
Fri Sep 26 19:38:03 2003
Date: Fri, 26 Sep 2003 19:35:09 -0400
From: Mark Rogaski <wendigo@pobox.com>
In-reply-to: <76711.1064559296@verdi.nethelp.no>
To: nanog@merit.edu
Mail-Followup-To: Mark Rogaski <wendigo@pobox.com>, nanog@merit.edu
Errors-To: owner-nanog-outgoing@merit.edu
--tKW2IUtsqtDRztdT
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable
An entity claiming to be sthaug@nethelp.no (sthaug@nethelp.no) wrote:
:=20
: > When an ISP buys a router does it want a worst-case guarantee about t=
he
: > router's capabilities? Or will it buy a router which can give better
: > performance in the average case (it may drop some packets if the traffic
: > pattern changes suddenly)? Assuming both cost the same.
:=20
: Worst case guarantee is necessary in many cases. Easy example:
:=20
: A router that can handle an STM-1 of regular Internet traffic is worthless
: to us if it dies in the face of an STM-1 with minimum sized attack traffi=
c.
:=20
Perhaps we can generalize this by pointing out the dearth of SLA's based
upon average-case. =20
Mark
--=20
[] Mark 'Doc' Rogaski | Guess what? I got a fever! And the only
[] wendigo@pobox.com | prescription ... is more cowbell!
[] 1994 Suzuki GS500ER | -- Christopher Walken (as Bruce Dickinson)
[] 1975 Yamaha RD250B |
--tKW2IUtsqtDRztdT
Content-Type: application/pgp-signature
Content-Disposition: inline
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (GNU/Linux)
iD8DBQE/dM0t79qnkXHz9aMRAhXiAKCAxjCdYhTfEtHgXhD63LYBH8MB3QCfWPz8
IjS9Ftz2XjtPI6rYE6OPJJg=
=ZxS9
-----END PGP SIGNATURE-----
--tKW2IUtsqtDRztdT--
