| Working Group:
| Revision Number:
IPv6 Working Group Minutes, RIPE58
Date: Tue, May 5, 2009, 16:00-18:00
Chair: David Kessens
Minutes: Vesna Manojlovic, RIPE NCC
Jabber monitor: Fergal Cunningham, RIPE NCC
Transcript available at:
Webcast available at:
A. Administrative Stuff
David Kessens (WG chair) went through the agenda:
B. Developments/Initiatives Regarding IPv6 in the RIPE Region and Beyond
B1: IPv6 peering BoF
Paul Hoogsteder, Breedband Delft (community network)
Paul announced that an informal session between old and new IPv6 peers
would take place on Wednesday, 16:00 near the cloakroom.
B2: Rumy Kanis, RIPE NCC, Training manager
Rumy made a brief announcement that the RIPE NCC Training Department is busy recording testimonials from people in the community who have hands-on experience deploying IPv6. She asked if anyone is interested in being filmed and interviewed about their experiences with IPv6, they should contact training _at_ ripe _dot_ net
C. RIPE NCC IPv6 update
Erik Romijn, Information Services
Erik announced that all of the RIPE NCC's external services have been available over IPv6 since January 2009.
Lorenzo Colitti (Google) mentioned that one service does not support IPv6 - the socks proxy. It's internal only though.
Erik said that he was only referring to public services.
Patrik Falstrom asked if ROSIE (the RIPE Meeting website) counts as a non-external service?
Erik responded that all ROSIE content can be accessed over IPv6 through the www.ripe.net web page, so you can actually see everything on ROSIE without having IPv4.
Mohsen Souissi, AFNIC, asked how do you think you can motivate the operators managing the reverse delegations for IPv6 to get the transfer active for you so that your Hostcount++ will work in IPv6? Do you have any strategy to get it better than zero?
Erik said that the first step is to get a zone transfer from a TLD.
Mohsen said that the first step is reverse delegation.
Andrei Robachevski, RIPE NCC, said that, for IPv6, we do not reverse DNS. We have to do this for IPv4. Because we can't transfer all the forward zones, we are just numerating all reverse zones. We can't do the same for IPv6 because IPv6 reverse DNS is big and sparse. So, the real problem here is that we have difficulty in actually transferring forward zones and therefore, we have difficulty counting IPv6 addresses there.
He continued that the RIPE NCC's strategy is to convince operators to open the Access Control Lists (ACLs) for RIPE NCC IP addresses the zones can be transferred.
For those operators who have some privacy issues or other considerations, we have developed a do-it-yourself kit to give to ccTLD operators. This software will count the hosts, anonymise the data and ship data back to us.
An attendee asked that, since this is related to forward DNS, do you think it's a good idea to contact ccTLDs?
Andrei responded that the RIPE NCC is using this idea but it's not 100% successful.
David Kessens asked if there was an explanation for the huge number AAAA requests?
Anand Buddhdev (RIPE NCC). Those quad-A queries are anomalous. They originate from broken versions of BIND. We suspect somebody is abusing them to send queries to the root.
Gert Doering asked what about the RIPE NCC's internal network and the workstations? Macs are v6 capable. Are you keeping it on, between people working there and your internal services, or are you afraid of that?
Erik said that on the RIPE NCC's wireless network, IPv6 is switched on and there are no problems.
Gert said that that was his experience too: it just works.
D. Google IPv6 update
Lorenzo Colitti, Google
Lorenzo explained that public IPv4 address space is going to go away. There is a business case for IPv6 because the alternatives are expensive.
There is also an opportunity here - there is going to be a large pool of devices that will only be accessible and able to access using IPv6. If you are a content provider, then you might want to talk to them which means you need to have IPv6.
It took google about a year-and-a-half to go from zero to being able to serve most services over IPv6. And we keep adding new ones. And with a very, very small team - just a handful of people.
He continued, saying that if you are an exchange point operator or if you are an ISP, please do not charge separately for IPv6. Please absorb this cost in other ways because charging separately will hinder adoption and users and traffic can appear overnight.
Extra data - measuring IPv6 query requests over IPv6:
0.25% of the total is over IPv6.
0.10% is over native, or not tunneled, IPv6.
2/3 is the proportion of requests that we serve as Google over IPv6. So two thirds of native IPv6 Internet appears to be enrolled in "Google over IPv6" program.
Mat Ford (ISOC) commented that the "Google over IPv6 program" does not scale. He asked at which number of your users who would be affected by not being able to reach www.google.com over IPv6 do you say "I don't care"?
Lorenzo answered that he did not don't but the magic number seems to be 10%. It depends on how situation evolves. We need to see better data on the "brokeness". It's hard to measure.
Kostas Zorbadelos (OTC Greece) asked do you use the same load-balancing mechanism in IPv6 for IPv4.
Lorenzo answered that Google tries to use the same infrastructure for both IPv4 & IPv6. The load-balancers are the same. Deploying separate hardware for IPv6 is not a good idea.
Jasper Wonnink (fix6.net, via Jabber)asked when will the Google bot start crawling over IPv6? When there is AAAA available?
Lorenzo stated that Google is working on this.
CK Claffy said that he objected to Lorenzo's statement "Just because we can't see the traffic, that doesn't mean that IPv6 is not really happening". As a scientist, the traffic is the only measure that IPv6 is really being used.
He said he knows how hard it is to measure the traffic and to get the traffic numbers but we should not be believing PowerPoint slides. This has has led us into a global telecommunications crisis, because we believed in numbers like "traffic is doubling every 90 days". That's a global example of how important is it to get traffic numbers, and what happens if you don't.
He continued that the EU is going to conduct a survey. He said he wanted to make it clear that a survey is a substitute for the truth. It's not actual data.
Lorenzo responded that CK was correct. He said that what he meant was do not expect smooth growth of IPv6 traffic. We are moving traffic from IPv4 to IPv6 because the user stacks, all else being equal, will prefer IPv6 to IPv4.
KC said that he thought it would be good if Lorenzo's graph about IPv6 traffic had a Y axis with numbers and labels so that I could tell it wasn't going from 0 to 0.002.
Lorenzo noted this.
Gil Mason (Altna, via Jabber) asked how Google defines production quality? When is a network good enough to to be allowed to connect to Google over IPv6?
Lorenzo responded that when we have a good connectivity and when the network is well supported. We like to see that the infrastructure is common, so that we don't get IPv6-specific failures.
Randy Bush (IIJ) said that measurable spikes can be seen, as large customers change something - the way DNS resolves, client roll-outs, gateway services etc. He continued that it depends on what kind of electronic microscope you are using. There's very little traffic. Google's measurements are as useful as any other I've seen.
Lorenzo said that 0.2% is coming over IPv6. There are daily patterns, not only spikes.
KC added that there are no *numbers* in this data. we need to acknowledge what is the limitation of the data we are making available and what data we would need to answer the question for real.
Lorenzo answered that client penetration is also a reasonable measure, or at least a measure we want to track.
Mohsen Souissi said congratulations but the presentation is not serving the community because, if taken out of context, it does not make clear that this is just one of the measurements tools, and that you do not see all the traffic, nor other servers that get deployed every day.
Lorenzo: These numbers only reflect IPv6 clients connected to our web site.
Jasper asked if Google would roll out IPv6 for www.youtube and when?
Lorenzo said that Google was working on all services but there are limitations in the hardware developer's time. It will come at some point.
E. Report(s) about *actual* v6 traffic volume as compared to v4? *what's real* out there, not what's on Powerpoint? (input from the audience)
Martin J. Levy,n Hurricane Electric
Martin explained that real traffic numbers are available in my slides from the Google IPv6 conference.
He continued that in his network, they do not have the same customer base on IPv4 and IPv6. Our stats about IPv6 can not be compared with IPv4.
I want to show 6to4 and Teredo traffic. 6to4 is not a bad protocol.
Teredo is a horrible protocol.
On our backbone, we introduced several 6to4 relays, and localizing traffic made a big difference. It peaks at 20 Megabit.
On Teredo, we were seeing hundreds of megs of Teredo traffic moving around. We deployed a Teredo service over the whole network, from 3 to 13, geographically dispersed. We were peaking around 150 megabits. Now we have 200-300 Megs of predominantly end user to end user traffic.
The easiest way to get rid of the Teredo, is to enable IPv6. It may be a great measure of how many people want to access v6 but don't have it.
An attendee asked how can we get rid of the horrible Teredo protocol?
Martin answered that running stable relays will get us though this transition period.
Randy Bush (IJJ) said that the optimist thinks the glass is half full; the pessimist thinks the glass is half empty. The engineer knows the glass is too big.
Martin answered the majority. We had to run 6to4 on every Teredo relay, on the same box, or next to each other.
An attendee commented that he does that with a dedicated connection between them.
F. Global IPv6 routing table status
Gert gave the routing table report. He asked are we growing fast enough? and said no, we are not.
David Kessens said that the advantage is that we have time to talk about that issue at the next meeting.
H. Overview of ISOC's current activity with regard to IPv6 deployment
Matt Ford, ISOC
I. Proposed EU IPv6 survey
Maarten explained that there was a commitment from the government to help make this happen, but for them to know what to do, they need to know where they can help the best.
He said that the European Commission has contracted TNO to carry out deployment monitoring activity. There will be a survey conducted in June.
He said that a draft survey is available for feedback.
The results on an aggregate level will be public, and will be used for information to the European Commission on their policy preparation.
Please contact maarten @ gnksconsult.com for more information.
Mohsen Souissi (AFNIC) asked if there is any initiative to harmonise all those various measurements to have common understanding about what do we measure, what the methodology is etc?
The tool was published at the last RIPE meeting. There may be many others. I feel frustrated, because there is no overall picture about who measures what.
Leo Vegoda (ICANN) asked if Maarten meant RIPE NCC members only, or the 15,000 other organisations that have resources from RIPE NCC, but are not members?
Maarten responded that he also meant non-members; anyone in the RIPE community.
Leo said the RIPE Community are not the only people who are operating networks in Europe. Will you take a list from ARIN too?
Maarten said no, the scope of this survey is limited.
David Kessens made a few final comments: we are an operators' group, a technical forum, but it's also important to make sure that policy makers, decision makers and governments are well informed about the challenges that we are going to face in the next few years regarding this IPv4 exhaustion.
He added that this is one of those projects that can contribute tremendously there, because it's basically an interface between those two worlds. So, I would very much like to encourage people to fill out the survey.