[178428] in North American Network Operators' Group
Re: Verizon Policy Statement on Net Neutrality
daemon@ATHENA.MIT.EDU (Tom Taylor)
Fri Feb 27 15:27:55 2015
X-Original-To: nanog@nanog.org
Date: Fri, 27 Feb 2015 15:18:06 -0500
From: Tom Taylor <tom.taylor.stds@gmail.com>
To: William Herrin <bill@herrin.us>, Scott Helms <khelms@zcorum.com>
In-Reply-To: <CAP-guGX=6-0PP9ozuWZu6BceZfsMMEHfk+fWU8Sq5TKowWzhXQ@mail.gmail.com>
Cc: NANOG <nanog@nanog.org>
Errors-To: nanog-bounces@nanog.org
On 27/02/2015 2:50 PM, William Herrin wrote:
> On Fri, Feb 27, 2015 at 2:22 PM, Scott Helms <khelms@zcorum.com> wrote:
>> I have to take exception to your example.
>>
>> Water, gas, and to a great extent electrical systems do not work on
>> oversubscription, ie their aggregate capacity meets or exceeds the needs of
>> all their customers peak potential demand, at least from "normal" demand
>> standpoint.
>
> Hi Scott,
>
> Do you propose that Internet access service should NOT be expected to
> meet peak "normal" demand? That would certainly make ISP operating
> models unique among public utilities.
>
> Regards,
> Bill Herrin
>
I've worked on both data network (Canada's X.25 Datapac) and
circuit-switched network provisioning (Nortel's DMS switches, and some
of my contributions appear in the ITU-T Orange Book). Circuit-switched
provisioning had the useful concept of "grade of service". This meant
that you set a target probability of delay or loss for a given load
level on the network (Average Busy Season Busy Hour, 10 High Day Busy
Hour, separate targets for each and provision to the most binding).
The same general concepts surely apply to IP network provisioning: you
know you can't economically serve all the traffic at the absolute peak,
but you set reasonable targets, assure yourself by simulation and
analysis that your design will meet the target, and build accordingly.
Tom Taylor
>
>
