IPv6 Matrix Project

Olivier MJ Crepin-Leblond — Sep 19, 2012 10:55 AM
Filed under: ,
In this article we describe the IPv6 Matrix, a project tracking IPv6 Worldwide. It is especially interesting to see how World IPv6 Launch on 4 June 2012 has helped IPv6 deployment in certain countries.

Introduction

Two of the Regional Internet Registries (RIRs) have reached their last /8 of IPv4 address space which will be allocated based on special policies. The APNIC started allocating from their last /8 in April 2012 and the RIPE NCC has reached their last /8 on 14 September 2012. In the image below you can see the predictions on when the other RIRs will approximately run out of IPv4 (when you click on the image, you will get to the most up to date graph).

RIR IPv4 Address Run-Out ModelFigure 1: IPv4 Address Report as provided by Geoff Huston

During World IPv6 Launch on 4 June 2012, many network operators, web companies, home router vendor and other Internet companies agreed to switch on IPv6 permanently. We wanted to find out if this has actually worked and how it improved IPv6 deployment in various regions.

The IPv6 Matrix project

The Internet Society chapter in the UK (ISOC England) has been awarded a community grants award in November 2009 to design and implement an "IPv6 crawler", a software that would crawl through the DNS at regular intervals in order to detect and test the following:

  • IPv6 DNS servers
  • IPv6 compliant web servers
  • IPv6 compliant SMTP servers
  • IPv6 compliant NTP servers

Methodology

Today over 95% of Internet traffic is generated by a small number of web sites. That means without IPv6 accessible content, IPv6 has no chance of ever being used and deployed widely.

We took the 1 million most popular web sites from alexa.com as a starting point to identify those domains that should be tested. More domains were then added later. We tested if these domains are reachable over IPv6. Note that this is equivalent to testing about 6.3. million hosts worldwide. A GeoIP database was used to determine the location of each host.

What are we tracking?

In order to determine host IPv6 penetration under country-code top level domains (ccTLDs) and a number of generic TLDs (.com, .net, .org), we looked at two types of information:

  • Infrastructure required to run IPv6 services: DNS + Web + email+ NTP server
  • Actual web content (this usually shows a lower percentage than infrastructure)

We also tested connectivity to these services (in case IPv6 addresses were advertised but no service was running on them). We also traced the route between our server at the London Docklands to each host on the list, using both IPv4 and IPv6 (all this information is archived in text format). All that is providing a lot of data, accessible from the IPv6 Matrix web site archive. In August 2012 over a dozen full runs were completed, each run taking a month and a half to complete. The size of this database was approximately 160 Gb and it is still increasing (the testing software runs 24 hours a day).

Results

Trends in Europe

Luxembourg, Norway and Germany see significant rise in IPv6 enabled infrastructure since World IPv6 Launch. This is likely to be caused by one major hosting provider installing dual-stack name servers. Slovakia and Portugal are leading with the number of dual-stack web sites. The Czech Republic shows the highest growth in dual-stack web sites during this period. Germany is the country with the largest number of dual-stack web sites in Europe.

The image in Figure 1 shows IPv6 enabled infrastructure in Europe measured on 4 June 2012 (blue) and in August 2012 (red).

Europe InfrastructureFigure 1: IPv6 deployment in Europe (infrastructure)

 

The image in Figure 2 shows IPv6 enabled web sites in Europe measured on 4 June 2012 (blue) and in August 2012 (red).

Europe WebFigure 2: IPv6 deployment in Europe (Web only)

 

Trends in Asia

Singapore is still leading by far in terms of IPv6 enabled infrastructure, but is trailing in dual-stack web sites. Armenia shows a growth of 2.72% in IPv6 infrastructure within two months. China shows abnormally low results. It is not clear if this is caused by content not being dual-stacked or the fact that IPv6 is hidden behind a firewall. India also shows low results, both in IPv6 infrastructure and the most popular web sites. Please note that the results for Asia are less reliable due to a smaller sample size in many countries in that region.

The image in Figure 3 shows IPv6 enabled infrastructure in Asia measured on 4 June 2012 (blue) and in August 2012 (red).

Asia InfrastructureFigure 3: IPv6 deployment in Asia (infrastructure)

 

The image in Figure 4 shows IPv6 enabled web sites in Asia measured on 4 June 2012 (blue) and in August 2012 (red).

Asia WebFigure 4: IPv6 deployment in Asia (Web only)

 

Trends in Africa

Many countries in Africa have IPv6 capabilities and local infrastructure, some through tunnels. South Africa shows the fastest growth in IPv6 enabled infrastructure. It is also interesting to see some dual-stack islands appear on the continent.

The image in Figure 5 shows IPv6 enabled infrastructure in Africa measured on 4 June 2012 (blue) and in August 2012 (red).

Africa InfrastructureFigure 5: IPv6 deployment in Africa (infrastructure)

 

The image in Figure 6 shows IPv6 enabled web sites in Africa measured on 4 June 2012 (blue) and in August 2012 (red).

Africa WebFigure 6: IPv6 deployment in Asia (Web only)

The current development in Africa can be compared to the growth of Internet connectivity in the 1990s. In the two maps below you can see how many countries were connected in June 1994 (on the left side) and in May 1997 (on the right side). Full Internet connectivity is in yellow with red cross-pattern. UUCP dial-up only connectivity is in green. Please refer to the Internetology web site for more information on other continents and dates.

Internet Connectivity in AfricaFigure 7: Development of Internet connectivity in Africa

Trends in North and South America

The US shows a significant growth, both in IPv6 enable infrastructure and in the number of dual-stacked web sites. This seems to be directly related to the World IPv6 Launch. Also Chile shows significant growth (5.39%) in infrastructure. Canada is trailing in the percentage of dual-stacked web sites. Note that in other countries, the data is based on a low number of web sites (the restricted number of hosting providers can make figures jump several percentage points).

The image in Figure 7 shows IPv6 enabled infrastructure in the Americas measured on 4 June 2012 (blue) and in August 2012 (red).

Americas InfrastructureFigure 8: IPv6 deployment in North and South America (infrastructure)

The image in Figure 8 shows IPv6 enabled web sites in the Americas measured on 4 June 2012 (blue) and in August 2012 (red).

Americas WebFigure 9: IPv6 deployment in North and South America (Web only)

 

Worldwide Trends

On a global scale, we are still seeing a slow growth in the deployment of dual-stacked hosts. A decrease in percentage in some countries could be related to unstable peering arrangements (the IPv6 network is less densely meshed than the IPv4 network). The US has clearly shown the fastest growth in this period, most likely a direct result of the World IPv6 Launch. Some small countries like Luxembourg had their main ISPs start offering dual-stack and consequently showing some growth in dual-stacked infrastructure. Bearing in mind that the Regional Internet Registry for the Asia Pacific (APNIC) has already reached their last /8 earlier this year, it is alarming to see so few web sites running IPv6 in that region, especially in countries with very high Internet growth like India. The World IPv6 Launch seemed to have had no impact in those countries.

 

Possible Errors/Caveats

We detected many errors in the DNS (commas, no A or no AAAA records, looping MX records etc.). Sometimes firewall and security software was blocking network segments or they are detecting denial of service attacks by mistake (some firewalls detect "ping" as being a denial of service attack and they detect "traceroute" as a hacking attack - thus blocking all traffic from the source address.).

Also note that the results provide a snapshot from one location only. As mentioned above, we achieved less accurate results in places we could only test a small number of domains (not enough significantly large local Web sites by traffic). Finally, the accuracy of the GeoIP database is disputable.

 

Future Work

The current front end web pages are only an example of possible analysis. In the future we are hoping to:

  • Develop new data visualisation
  • Perform further analysis
  • Perform historical analysis and analysis over a longer period of time from archives
  • Develop an engine to write automated reports

We are also planning to duplicate the crawler software to other regions to create multiple views. You can find more information on the IPv6 Matrix web site. If you are interested in some more details or if you would like to support this project, please contact us at ISOC England <contact _at_ isoc-e _dot_ org>.

0 Comments

Add comment

You can add a comment by filling out the form below. Only plain text is possible. Web and email addresses will be transformed into clickable links. Comments are moderated so they won't appear immediately.

Related Items
Increased Reach of RIPE Atlas Anchors

Increasing the reach of RIPE Atlas anchors is one of the highest priority goals of RIPE Atlas Team. ...

Proposing Making RIPE Atlas Data More Public

RIPE Atlas is now three years old, and is moving from a prototype to production service. Based on ...

Modifications to the IP Analyser to Reflect New Policy

We are in the process of implementing the policy regarding Post Depletion Adjustment of Procedures ...

Report on IPv6 Security Test Methodology

The Dutch Institute for Applied Scientific Research (TNO) and a number of Dutch security companies ...

RIPE Atlas: Improved Probe Pages

We've made it much easier to get an overview of the history and measurements for all the public ...

more ...