A good scheme needs to have several attributes: -- a pricing model??

[David Farber <dave () farber net>]

________________________________________
From: Tony Lauck [tlauck () madriver com]
Sent: Monday, June 16, 2008 3:46 PM
To: David Farber
Subject: Re: [IP] a pricing model??

A good scheme needs to have several attributes:

1-Users should be able to understand how resource allocations are made.

2-Users should feel that resource allocation is fair, in the sense that
they are not suffering because of network "hogs".

3-Users should receive a predictable monthly bill.

4-Users should not be punished (e.g. disconnected) for "excessive" use.

5-Users should be encouraged to moderate their demand during congested
periods.

6-Users needing heavy usage should be able to pay more and get better
service.

7-Resources should never be idled for policy reasons.

9-Network operators should have predictable revenues and costs.

9-Implementation cost should be minimal to insignificant compared with
other costs.

10-Resource allocation and billing should scale with number of users,
network topology and available technology (e.g. link and router speed).

I believe these goals are practically achievable, at least where there
is a single bottleneck queue (so this would include the DSL case, but
not the DOCSIS 2.0 case). The following is a sketch of how this can be done:

Users purchase service in units and pay a fixed monthly fee accordingly.
Associated with each unit is a guaranteed minimum data rate (at least as
good as dial-up) and a peak rate. The guaranteed minimum data rate is at
least 100% provisioned by the network operator. A leaky bucket scheme
prioritizes packets from users below their guaranteed minimum. These are
all the highest priority. Associated with each user is a priority number
associated with that user for packets in excess of the minimum. For
these packets the resource is administered as a priority queue. (A low
priority number results in the user receiving preferential service.)
Priorities are updated frequently (at least every few seconds) as
described in the next paragraph.

Each packet by a user is counted as a contention packet if servicing the
packet has delayed another user. The effect is that heavy users who use
the network during busy periods will have very high counts, while light
users or heavy users who use the network during idle periods will have
low counts. Users are assigned priority queue number by dividing their
contention packet count by the number of service units they have
purchased. So that reformed network hogs are not unduly penalized, the
user's count can be exponentially aged, perhaps with a half life of one
  month.


Tony Lauck
www.aglauck.com



David Farber wrote:
> ________________________________________
> From: nelsberg () gmail com [nelsberg () gmail com] On Behalf Of Ted Nelson [tandm () xanadu net]
> Sent: Monday, June 16, 2008 3:36 AM
> To: David Farber
> Cc: Ted Nelson
> Subject: Re: [IP] a pricing model??
>
> It's fine to look for optimum pricing, but isn't
>  charging by bit rate a heck of a lot simpler
>  than an instantaneous global flow metric?
>
> Best, Ted
>
>
>
>
> -------------------------------------------
> Archives: http://www.listbox.com/member/archive/247/=now
> RSS Feed: http://www.listbox.com/member/archive/rss/247/
> Powered by Listbox: http://www.listbox.com



-------------------------------------------
Archives: http://www.listbox.com/member/archive/247/=now
RSS Feed: http://www.listbox.com/member/archive/rss/247/
Powered by Listbox: http://www.listbox.com