[141015] in North American Network Operators' Group
Re: Contention/Oversubscription maths
daemon@ATHENA.MIT.EDU (Scott Weeks)
Fri May 27 16:22:57 2011
Date: Fri, 27 May 2011 13:22:50 -0700
From: "Scott Weeks" <surfer@mauigateway.com>
To: <nanog@nanog.org>
Reply-To: surfer@mauigateway.com
Errors-To: nanog-bounces+nanog.discuss=bloom-picayune.mit.edu@nanog.org
-------- lists@memetic.org wrote: ------------
From: Adam Armstrong <lists@memetic.org>
I'm more interested in the levels of traffic that we will see consistently.
-----------------------------------------
>From experience, one thing's for sure. No matter how you end up estimating traffic levels, you should optimize it using empirical data. Make your estimates, set up the connectivity and watch the traffic. In some areas you will be able to have a larger over-subscription rate because the customers will have lower traffic level internet habits and in other areas the over-subscription rate will be less because you might have a pocket of power users.
To the point some have made that user's internet surfing habits (thus more traffic) don't change because of an increase in bandwidth, rather they change more gradually, is definitely the case for residential subscribers. I have seen folks spend way too much time trying to get it right from initial estimates and totally blow it.
There is no substitute for empirical data!
scott
