[LACNIC/Politicas] IPv4 HD-ratio

[Geoff Huston]

Firstly, my apologies for doing this response in english.

I looked over the report referenced below, and I'd like to expand a little 
on the technique I used to report on the likely IPv4 exhaustion date using 
the supposition that RIPE would adopt an HD Ratio based allocation metric.

The basic assumption in this model is that the exponential trend of growth 
in advertised addresses over the past 3 years will continue at the same rate.

The next assumption is that the ratio of advertised to unadvertised 
addresses will continue to decline at a linear rate - i.e. the unadvertised 
address pool will also grow at an exponential rate, but at a slower rate 
than advertised addresses.

The next assumption is that the relative allocation rates of each RIR will 
change over time in proportion to an exponential growth model of the 
previous 3 years allocation rates

For any day, the amount of addresses to be allocated equals the sum of the 
growth in advertised addresses and the growth in unadvertised addresses. 
Each RIR will allocate addresses in line with the relative allocation rates.

To simulate the adoption of an HD Ratio in RIPE, I used the same 
exponential growth model as before, but then added on top of this the 
incremental number of addresses as calculated by the relative difference 
between an 80% allocation and an HD allocation.  On average this is an 
additional 17% (45% of 38%), but the model uses a more complex algorithm. 
The unadvertised address count is based on the higher advertised count (I'm 
not sure that this should be the case, but equally I'm not sure that it 
should not!).

The model is driven forward in time as before, and I'm looking for the data 
where the IANA unallocated address pool exhausts.

In the model where all RIRs use the 80% allocation metric the model I've 
used predicts that the remaining 61 /8s are likely to be distributed in the 
following fashion:

AFRNIC: 0
APNIC: 24
ARIN: 9
LACNIC: 3
RIPE: 25


In the model where RIPE adopts the HD Ratio immediately, then the IPv4 
address pool will exhaust 13 months earlier, and at that point in time the 
remaining 61 IANA /8s are distributed, according to this model as follows:

AFRNIC: 0
APNIC: 21
ARIN: 9
LACNIC: 1
RIPE: 30

There is a timing issue here between LACNIC and APNIC, and its also 
possible that the outcome could be that the remaining /8s are distributed 
along the following lines:

AFRNIC: 0
APNIC: 20
ARIN: 9
LACNIC: 2
RIPE: 30

Of course, don't forget that these are just predictions from one form of 
model, and there are other ways to model this situation, and, in any case, 
reality always seems to find its own way forward!

regards,

     Geoff




At 08:05 AM 24/02/2006, Oscar A. Robles-Garay wrote:
>Lo pueden ver en
><http://docs.nicmxlabs.org.mx/hdratio/hdratioipv4.pdf>http://docs.nicmxlabs.org.mx/hdratio/hdratioipv4.pdf 
>
>(45 KB)
>Saludos,
>
>Oscar
>
>At 02:59 p.m. 23/02/2006, Oscar A. Robles-Garay wrote:
> >Parece que el servidor de LACNIC no dejó pasar el attachment... va de nuevo.
> >Oscar
> >
> >At 02:56 p.m. 23/02/2006, Oscar A. Robles-Garay wrote:
> >>To: "rgaglian at adinet.com.uy" <rgaglian at adinet.com.uy>,
> >>         <politicas at lacnic.net>
> >>Subject: Re: [LACNIC/Politicas] IPv4 HD-ratio
> >>
> >>Roque,
> >>
> >>Justo esa es mi preocupación.
> >>Hicimos un ejercicio con los datos de
> >>http://ipv4.potaroo.net aquí se los anexo,
> >>tiene algunas inferencias (ahí mismo las explico).
> >>Las conclusiones son importantes:
> >>         - Nos acabamos el espacio de direcciones más rápido
> >>         - Va a haber menos direcciones para
> >> los demás RIR's (10 /8 menos, en el mejor de los casos)
> >>
> >>Creo que debemos hacer algo al respecto.
> >>
> >>Oscar
> >
> >
> >_______________________________________________
> >Politicas mailing list
> >Politicas at lacnic.net
> >http://www.lacnic.net/mailman/listinfo/politicas
>_______________________________________________
>Politicas mailing list
>Politicas at lacnic.net
>http://www.lacnic.net/mailman/listinfo/politicas