Slashdot: UK ISP PlusNet Testing Carrier-Grade NAT Instead of IPv6
owen at delong.com
Thu Jan 17 23:50:03 UTC 2013
> Nevertheless, I'll be happy to document my assumptions and show you
> where they lead.
> I assume that fewer than 1 in 10 eyeballs would find Internet service
> behind a NAT unsatisfactory. Eyeballs are the consumers of content,
> the modem, cable modem, residential DSL customers.
And this is where you run off the rails… You are assuming that NAT today
and CGN provide similar functionality from an end-user perspective.
The reality is that they do not. CGN is a substantially more degraded
form of internet access than current traditional per-site NAT.
1. The end-site does not control the NAT box.
2. UPnP and NAT-PMP do NOT work through CGN.
3. There is no other provision in most CGNs to allow for inbound
connection trickery that allows many of today's applications to
function in spite of NAT.
> Some few of them
> are running game servers, web servers, etc. but 9 in 10 are the email,
> voyage and netflix variety who are basically not impacted by NAT.
Vonage will, in most cases fail through CGN as will Skype, Xbox-360,
and many of the other IM clients.
> I assume that 75% or more of the IPv4 addresses which are employed in
> any use (not sitting idle) are employed by eyeball customers. Verizon
> Wireless has - remind me - how many /8's compared to, say, Google?
Are you sure that 75% of VZW's IP addresses are assigned to end-customer
devices? I am not.
> If you count from the explosion of interest in the Internet in 1995 to
> now, it took 18 years to consume all the IPv4 addresses. Call it
> consumption of 1/18th of the address space per year.
I'll leave the obvious math error in this assumption as an exercise for
>> From my assumption, 25% of the addresses are consumed by non-eyeball
> customers who will continue consuming them at 1/(18*4)= 1/72 of the
> address space per year. Assuming that server ops still need that many
> addresses when acquiring them is not so close to free.
This assumption ignores non-customer use of addresses which, while minor,
is not insignificant.
>> From my assumptions 75% * 0.9 = 67.5% of the addresses are currently
> consumed by eyeball customers who can convert to NAT. Match the
> previous paragraph's math at 49/72's of the address space recoverable
> at some cost that while not trivial is also not exorbitant.
This makes a rather absurd assumption that the majority of those eyeball
addresses are not already assigned to eyeball NAT pools. This is the
second place where your assumptions run wildly off the rails IMHO.
> Eyeballs were consuming at (1*3)/(18*4)= 3/72's per year but if only 1
> in 10 needs a global address that slows to 3/720's.
While the math works, it would be a lot more clear to say 1/4 * 3/18 = 3/72.
> 13/720's per year consumes 490/720's after 37 years.
> 37 years.
> So, where am I wrong? Is it more like 1 in 5 customers would cough up
> an extra $5 rather than use a NAT address? The nearest comparable
> would be your ratio of dynamic to static IP assignments. Does your
> data support that being higher than 1 in 10? I'd bet the broad data
> sets don't.
First, it's more like 1/100 customers that are not already behind NAT
of some form, so your 37 years drops to 0.37 years (a little more than
This seems very disruptive and rather heavy on the overhead for a 4-month
More information about the NANOG