RE: detour lives...

Stefan Savage (savage@cs.washington.edu)
Fri, 30 Apr 1999 20:33:37 -0700

This is true, on the other hand they aren't limited to a single IP address
(they can a pool of them).

- Stefan

-----Original Message-----
From: Kenichi Ishikawa [mailto:ishi@cs.washington.edu]
Sent: Friday, April 30, 1999 8:29 PM
To: savage@cs.washington.edu
Cc: syn@cs.washington.edu
Subject: Re: detour lives...

The thing I worried about NAT is the limitation of number of ports.

Suppose NAT could use 32768 ports, and 2MSL == 60 sec,
32768 / 60 = 546 (flows / sec)
Suppose 1 flow == 10 packets, 1 packets == 1500 byte
MaxBW = 32768 / 60 * 10 * 1500 * 8 = 8 Mbit/s

is this enough as a router ?

At Fri, 30 Apr 1999 20:04:25 -0700,
Stefan Savage <savage@cs.washington.edu> wrote:
>
>
> I have found confirmation that one or more web hosting agencies (DIGEX)
> sometimes use NAT's to pick the return exit-point. They put a NAT at each
> peering point and transform the source address on ingress so the response
> returns the "right" way. Pretty amazing kludge...
>
> - Stefan