Re: [BEHAVE] Fwd: New Version Notification for draft-nishizuka-cgn-deployment-considerations-00.txt

[Simon Perreault <[e-mail]>]

Le 2013-04-03 20:57, Senthil Sivakumar (ssenthil) a écrit :
> If it is not the address, what is the limiting factor? The reason ISP
> is deploying CGN is the shortage of addresses and cant provide a
> single address to each of his subscribers. Maybe you meant to say the
> address is not the only limiting factor. I never said the text was
> saying static isnt good enough :-), I deduced from the study that the
> usage of ports is far more compellingly efficient with dynamic port
> allocation and the cost of logging infra can be justified.

I'll try to illustrate my point with an example with numbers.

An ISP is running out of addresses. It considers two options: static CGN 
vs dynamic CGN. Static allows, let's say, 32 users per public IPv4 
address. Given the 1:10 figure from the draft, it follows that dynamic 
allows 320 users per public IPv4 address.

If 32 is "enough", why suffer the trouble of logging (among others) just 
to get to 320? If 32 and 320 are both "enough", then considerations 
other than efficient use of public IPv4 addresses must take priority.

"Enough" could mean something like "enough to support projected growth 
for X years".

> Most of the studies in the past projected how bad the logging problem
> is but didn’t have any data on the other side of the equation on how
> inefficient the static port allocation is. I wouldn’t want this draft
> to say one is better than the other, but let the operators choose, if
> 10:1 static to dynamic port allocation is justified for their
> deployment.

The 10:1 figure is useful. However, the conclusion "therefore dynamic is 
better" is premature. There are tons of other criteria to consider. Even 
worse, it is very possible that the 10:1 figure does not even matter: if 
static is "good enough", you may not care that dynamic is 10 times better.

Simon
-- 
DTN made easy, lean, and smart --> http://postellation.viagenie.ca
NAT64/DNS64 open-source        --> http://ecdysis.viagenie.ca
STUN/TURN server               --> http://numb.viagenie.ca