Google and IPv6

[Mike Leber]

On Sun, 16 Mar 2008, Erik Kline wrote:
> An additional concern would, of course, be performance (i.e. latency).

FWIW, I've been keeping track of the relative performance of v4 vs v6
using IPv6 reverse DNS nameservers (as an initial test set of hosts with
v4 and v6 addresses) and IPv6 performance has been improving steadily over
the last few months (a few percent a month).  As listed at
http://bgp.he.net/ipv6-progress-report.cgi

(as of Sun Mar 16 01:33:43 PDT 2008)
Percentage of IPv6 rDNS Nameservers where IPv6 is as fast or faster than
IPv4 (within 1ms): 53.4%

It is also worth noting that in some cases IPv6 is faster than IPv4.  
Differences in performance between IPv4 and IPv6 is due to different
network topology, which is causes by varied rates of implementation, which
means that peering relationships are fundamentally different (you can't
peer via v6 if you don't even run it).

I'm guessing this measurement will never quite reach 100 percent, however
some some set of networks IPv6 may end up being usually faster than IPv4,
which has some interesting economic implications (i.e. applications could
test and decide to use v4 or v6), or people that spend money could test
and decide).  Aside from what being faster means or doesn't mean, another
important benchmark is "as good as", hence why I am measuring "as fast or
faster".

I'll be adding other v6 deployment metrics soon.

Mike.

+----------------- H U R R I C A N E - E L E C T R I C -----------------+
| Mike Leber         Wholesale IPv4 and IPv6 Transit       510 580 4100 |
| Hurricane Electric     Web Hosting  Colocation                 AS6939 |
| mleber at he.net                                           http://he.net |
+-----------------------------------------------------------------------+