V6 still not supported (was Re: CC: s to Non List, Members, (was Making Use of 240/4, NetBlock))

Abraham Y. Chen aychen at avinta.com
Mon Mar 14 18:05:19 UTC 2022

Hi, Fred:

0)    Thanks for sharing your references to IPv6 statistics.

1)    However, you might have looked this topic too deeply and missed 
the overview. Through our study of EzIP, we have discovered two aspects 
of this topic that frequently mislead readers:

     A.    IPv6 equipment */Capability/* vs. IPv6 traffic */Volume/*:    
The former is about equipment "readiness". The latter is about the 
actual traffic "amount". High value of the former does not necessarily 
imply the same to the latter. Without explicitly stating which one is 
being presented, wasteful debates based on these numbers persist. In 
fact, this is where the big smoke screen that our team had to go through 
to finally see more clearly about the true reference that we have to 
base upon.

     B.    The scope of the candidates for the source data set:    
Comparisons of the incident type needs to find the largest domain, not 
several sub-domains, to extract data from. This is important because an 
author can pick up a few datasets that support the "theory", while 
overlooking others that may be more significant and introduce counter 
views. This is also tricky, because how can one tell which type of 
domain is bigger than the others? Since we are discussing a global 
event, any domain that implies worldwide coverage is much more 
trustworthy than multiple full datasets, each is country-based .

2)    Although all of your citations contain "IPv6", some even include 
"status", none of them clearly defines which one of the two choices in 
Pt. 1) A. it is referring to. Except, the browser tab of the following 
reads "IPv6 */Capability/* Metrics". This confirms, at least to me, that 
you have been looking at Capabilities instead of Volume. The former 
tends to show much higher numbers because it is just a report of how 
many devices online are ready to use IPv6 (I heard that such indication 
is carried in the IP packet somewhere?). This is because when users 
actually use IPv6, it then contributes to the Volume statistics.


3)    Allow me to share with you the IPv6 statistics that we have been 
following. On top of the two criteria in Pt. 1) above, we also look for 
consistency through time, i.e., track records such as how long and how 
frequently the data is updated.

     A.    IPv6 Deployment / Readiness: The following has been updated 
every few days for quite sometime. Whatever metrics that it is based 
upon, we can assume that it is applied worldwide. This appears to be a 
superset of what you cited.


     B.    IPv6 Traffic Volume: There was an annual report series by 
Cisco. But, the last one was in 2017. We could not find any later 
versions. Interestingly, Cisco TechSupport confirmed so.


4)    These days, we primarily monitor the following two statistics to 
keep up with the IPv6 performance status:

     A.     Google IPv6 Adoption: This is a daily report that goes back 
to 2008 Oct. The current peak is around 38%, while the average is around 
36%. Also, a very interesting phenomenon can be observed if you zoom 
into a small segment of the graph. It shows the weekly fluctuations that 
peak during weekends or holidays when users are mostly accessing Google 
services from homes. In fact, the lock-down during the COVID-19 pandemic 
pushed the average up notably. Although this is only within Google, it 
is a worldwide statistics. More importantly, Google is one of the 
stronger IPv6 promoters. Similar statistics from other business should 
logically be capped by this graph.


     B.    AMS-IX Traffic Statistics:     There are various measurements 
that are constantly (between every 7 to 15 minutes) updated. The URL 
below leads you to a set of composite graphs showing the percentages of 
various type of protocols being transported through AMS-IX. What you 
will find in the below graph is that IPv6 Traffic is only around 4.5% 
which is in the 10% range of that in the Google graph above.


5)    Pt. 4) B. is a surprise statistics. We have not been able to 
pin-point exactly how could this happen. Two factors should be kept in mind:

     A.    AMS-IX may not be the largest IX (Internet eXchange), but 
their operation is worldwide. They are the only business of this kind 
that has been providing the continuous reports (This kind of reports 
have been going on for ages. AMS-IX does not provide archived data. But, 
you can search Archive.org for historical data as far back as 2010-06.). 
So, we believe that this is more reliable than others.

     B.    IX businesses take the overflow traffic from Internet 
backbone carriers' peering arrangements. So, the ratio between IPv4 and 
IPv6 could be different from the bulk in the core. However, one 
historical event (see URL below) hints that the IPv6 traffic on AMS-IX 
should have been even lower than what is reported if its peering 
agreements had been settled in a similar manner as those for IPv4.


Hope this run-down of background and history enable us to synchronize 
our perspective of the IPv6 status.


Abe (2022-03-14 14:04)

NANOG Digest, Vol 170, Issue 15
Message: 15
Date: Sun, 13 Mar 2022 21:06:51 -0700
From: Fred Baker<fredbaker.ietf at gmail.com>
To: Joe Maimon<jmaimon at jmaimon.com>, "Chen, Abraham Y."
	<AYChen at alum.mit.edu>, "Abraham Y. Chen"<aychen at avinta.com>, Ca By
	<cb.list6 at gmail.com>
Cc: NANOG<nanog at nanog.org>
Subject: Re: V6 still not supported (was Re: CC: s to Non List
	Members, (was Re: 202203080924.AYC Re: 202203071610.AYC Re: Making Use
	of 240/4, NetBlock))
Message-ID:<7E0B159F-43AC-4507-8FB9-3120B2B9046A at gmail.com>
Content-Type: text/plain;	charset=us-ascii

> On Mar 11, 2022, at 8:39 AM, Joe Maimon<jmaimon at jmaimon.com>  wrote:
> Google's statistics...

I'm not sure which of you I'm replying to. The comment was made on NANOG the other day that we should discount Google statistics because they have been promoting IPv6 for a decade. It's true that they have been doing so. But they aren't the only people with statistics.


You might look at the following links. Eric Vyncke has been putting up charts basically on Google, Akamai, and APNIC statistics for a while. One thing to consider is that around 90 countries (92 in this capture, as low as 89 a couple of days ago) have 5% or greater response rate using IPv6. Google and Akamai have their own content networks, and in at least some countries only externalize AAAA records or respond to IPv6 requests. APNI isn't that way; they don't operate a content network, but rather accept traffic from across the backbone. Consider that a content network essentially reports traffic from a customer network to their first hop ISP, while when APNIC reports an IPv6 access, the father form APNIC to the collector in question has to include every network and every router in the path. Now look at these:


I think the APNIC numbers demonstrate that paths through the backbone generally support IPv6 end to end, and that from a routing perspective there is no reason to favor IPv4.

There are 8 Countries (this evening) that Google reports roughly equal response rates from using IPv4 or IPv6. cfhttps://www.vyncke.org/ipv6status/compare.php?metric=p&countries=in,my,sa,be,de,fr,gr,vn. This doesn't prove that IPv6 has taken over the world, but it does prove that those who would discount available statistics sources are a little too shrill in doing so.

Where IPv6 has a problem today is with enterprise. IMHO, this is basically because enterprise is looking at the bottom line. If ISPs were to do what Mythic Beasts says they do, which is charge their users for address space, IPv6 is virtually free while IPv4 costs something. I suspect that enterprise would change its tune dramatically.


This email has been checked for viruses by Avast antivirus software.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.nanog.org/pipermail/nanog/attachments/20220314/1d423f70/attachment.html>

More information about the NANOG mailing list