04 November 2008, 16:13

Google routes around the current black spots in the IPv6 internet

Search engine giant Google is taking the next steps towards IPv6. Admittedly, as Google developer and IPv6 expert Lorenzo Colitti said at the RIPE meeting in Dubai, having an IPv6 entry for the central www.google.com web page is still regarded as too ambitious. The problem is that about one in 10,000 users won't be able to access the Google page anymore if there is a dual-stack IPv4/IPv6 solution. Colitti said, although the numbers Google established in a study carried out with IPv6 users PDF were lower than expected, "But that's still not acceptable,".

Instead, he advocates as many direct peerings as possible to free up the connection to www.ipv6.google.com. In order to enable peerings with small partners, Google wants to include as many partners as possible into its own IPv6 peering table by making automated announcements via the "BGP community" tag PDF . While this tag is still being described as a "hack" by many experts as well as Colitti himself, it was received positively at the RIPE meeting.

According to Colitti, overly long response times in the traffic between IPv6 sites are one of the central problems that are delaying the widespread adoption of the new address protocol. Colitti described the vicious circle "The low adoption generates limited data traffic. Low data traffic leads to bad connectivity, and bad connectivity hampers a more widespread adoption". He said backbone and transit providers are currently doing a bad job with providing IPv6 connectivity. Long paths, less than optimal routing and bad middleware boxes cause the latency to be higher than with IPv4. The developer said that in some tunnels, traffic disappears never to be seen again. Google's researchers found in their study that an IPv6 packet needs 413 milliseconds from the US West Coast to China. Colitti said "We don't want to do this to our users".

To make the services it already supports via IPv6 accessible via IPv6, Google plans to cut any dysfunctional routes and peer with the IPv6 users themselves. Colitti said "We don't use these routes, and we don't take transit","Better no connectivity than bad connectivity". As he already did at the RIPE meeting in Berlin, Colitti invited peers to exchange IPv6 data traffic directly with Google. According to the developer, peerings have already been arranged with many of the large IPv6 providers and Google is, for example, already discussing this with French provider Free, one of the largest IPv6 providers.

To allow as many IPv6 users as possible access to Google's IPv6 content, small networks will be able to link into Google's network of "trusted users" via the "BGP community tag" (RFC 1997). Colitti said that for this purpose, the IPv4 resolver prefix of a participating network needs to be given a dedicated "community" tag like 15169:6666, , which enables these networks to be white listed automatically. The Google expert said that by using the corresponding community tag, providers signal their commitment to provide good IPv6 connectivity and support their own users when difficulties arise.

Google's persistent commitment to promote the new protocol was rewarded with a lot of applause from address managers. The initiative is regarded as an attempt to resolve the IPv4 successor's chicken and egg scenario. Although there will be no more new IPv4 addresses in three years' time, ISPs are waiting for "killer apps" while content providers are waiting to see demand from users first. Colitti was also applauded for calling on router manufacturers not to charge extra for IPv6 as an additional feature. A surcharge like this will make the support of IPv6 the first thing to economise on, warned various RIPE members from businesses and universities.

On the impending shortage of IPv4 addresses and the efforts and transition scenarios for IPv6, see also:

Feature: