Kurtis's Blog

Today I am in Dublin, Ireland invited by INEX, the internet exchange in Ireland. This one of these signs of how closely the IXPs in Europe actually co-operate. I am invited to talk about the broadband build out in Sweden and also a bit about the history of Netnod. My slides can be found here, and I will already acknowledge now that not everyone will agree with the conclusions :-)

The INEX meeting yesterday had a very good mix of technical and policy presentations. It's always good to come out and see some of the other IXes, meet their members and see how they are developing.

o Radio spectrum, Kevin Kennedy ComReg

First our was Kevin Kennedy from the Irish regulator ComReg. He spoke about the use fot he radio spectrum in Ireland. Usage of the radio spectrum is high, and the number of point-to-point fixed licenses has grown significantly in the last years. This was apparently used as an alternative to when connecting cell towers but more and more by business connecting their various branch-offices. Which I thought was interesting. Mr Kennedy also went on to note that he value of the radio spectrum to the economy is high. Referred to the radio spectrum as a natural resource to Ireland. I believe he is right and that this is an often overlooked fact. We tend to concentrate on the value in terms of mobile operator licenses, but there are also other uses of radio spectrum in our day lives and the telecom world.

o Beyond the Yellow Brick Road, Niall Murphy, Google

Niall argued that due the the exhaustion of the IPv4 address, a trading market will emerge and will also be needed tobridge the gap between IPv4 and IPv6. He argued that we need to make it happen and make sure that it happens in an operational viable way while we work on deploying IPv6.

I don't disagree with him on this, but he started the presentation with outlining how he believes the run out of IPv4 will look like. He also said they had been doing simulations similar to Geoff Hustons. I pointed out that both their model as well as Geoff's modelling on an exhaustion date is based on the assumption that LIRs/RIRs actually see all requests for needed address space. I don't believe that to be true. For what it's worth we are out of IPv4 space. Most of this is hidden behind NATs.

o How Sweden has tackled the broadband era, Kurtis Lindqvist

I then when on to present the slides I posted here yesterday. It seems to have been well received. Some comments I got was that Ireland is around 7 years behind. Based on the discussions yesterday they might be a bit ahead as they have deployed fiber in many of the larger cities. But it's true that they are facing many similar challenges ahead. I am sure we will have plenty of reasons to come back to this.

o Donal O'Cearbhaill on managing 4,000 schools broadband connections

This presentation was pretty interesting, and I hope the slides will be made available. Their efforts to use and integrate open-source software network monitoring tools to be used to monitor these 4000 connections to a single platform sounded quite useful. There are several good and often used open-source tools out there like Cacti, Nagios etc, but from an operational point of view you really want a single interface to these tools. Can't wait to get the slides!

o Barry Rhodes and Eileen Gallagher gave INEX updates

INEX growth seems to be hampered by member and potential member mergers. We have seen similar phases of market consolidation also at Netnod. It's true that it will impact the budget for your IX negatively, but in the long run I believe that all have to win from more stable members. An additional strategy for membership increase is to go after members that are out of your immediate geographical area betting that market consolidation will not as easily happen in several markets at the same time. For me joining Netnod in 2002 this is less clear. A global market down-turn will affect everyone and lead to consolidations across the board. And I happen to believe we are getting to close to another one. This time though we might end up in a turmoil while we are still at a lack of capacity due to under-investment and to low operating margins. But that is for another blog post.

After the members meeting there where drinks and I was again reminded of how much better Guiness tastes in Ireland!

So coming back from Dublin I am passing through the new Terminal 5 at LHR. Despite all the scare I have heard it's actually a pretty amazing terminal. check in queues was non-existant and security was really fast (but this was at 11.00)! I had heard some rumours that there wouldn't be any wireless here but that seems to have been fixed. I get wireless more or less everywhere.

What does dissappoints me is - power. How can you build a brand new terminal and no power-sockets. The Starbucks on lower grounds does have power though! Phew.

I'll also make the observation that I am not sure how they where thinking when they built a grand new terminal but only have 5 toilet stands per toilet...

Today at the RIPE meeting I presented and idea that was originally proposed by Ruediger Volk from DTAG and put into writing by Randy Bush. I merely volunteered to present it at RIPE...

The idea is really simple. Network operators have the problem that they want to filter BGP peer prefixes and validate the received prefixes. Several proposal have been made to sign the IRR data based on the content of the RPKI. However, several of these are fairly complex, at least as compared to what Ruediger proposed. Ruedigers proposal was that the RIPE NCC, or any RIR for that mater, would build and announce a new IRR based on the ROA data in the RPKI. This data would have the advantage that if operators preferred this over other IRR data, where an object existed, they would have a trusted validation of the prefix filtering they put in place. The advantage of this over a lot of other proposals is that the new registry would just be another IRR publication point and existing tools could be used.

This week I will be at the RIPE meeting in Berlin. I personally find RIPE meetings some of the most valuable meetings in the industry. It's the opportunity to see interesting technical presentations, discuss operational experiences, meet vendors and perhaps most of all - meet customers. This week we at Netnod will also sponsor one of the social events on the Wednesday night. I hope to see all readers there!

I will try and post views on anything interesting as topics are discussed!

In Swedish newspaper E24 I read that the crisis management agency (KBM) will investigate the dependencies of government and government agencies on TeliaSonera as mobile phone provider. This after a massive failure last friday that made their customers unable to call for hours.

Now, how serious was this failure really? Of course, it's not to be taken lightly that the largest mobile operator network is out for hours. That certainly deserves attention and investigation. But for KBM to look at this as a dependency for the government? Let's think about what the consequences where. Well, I am sure it's annoying that they couldn't call their colleagues, that would be true for any work place. I am sure it's annoying that they couldn't use their mobiles to call out. Same for anyone. So far it doesn't strike me as something that would bring a national crisis. So what would the case be if there was a national crisis during an outage? I am sure that would hamper the governing process and make communication harder - but to be honest, if the national crisis management is built around the assumption that mobile networks are operational - I think we have other issues. If that is the case we have a crisis management system that is broken (which I happen to believe it is anyway, but for different reasons). So what would the remedy be? Well, give different agencies phones in different provider networks? But what happens then? They can still call each other if they have the right provider....but to be sure you would not put all "critical" agencies in the same provider - and you still have gained very little as it's 'very' hard to predict who will be the network that fails (kind of the nature of the problem).

Other alternatives? Well, you could give each of the employees two phones, from different providers. Or somewhat more modern, one with two sim cards (but they are quite hard to come by). If we ignore some practical problems with that, we also have the problem. that many of the cell towers are actually shared. Both just the tower, but for cost cutting reasons several of the 3G operators (I am now assuming the government would use 3G. I know, silly me) have built networks together where they are both virtual operators inside the common infrastructure. So if that common infrastructure fails, you need to make sure you don't have overlap. It's starting to get very complex.

In the end, what was the problem we where trying to solve? At the parliament hearing I attended some time ago, someone made the point "The country tends to work quite well even if we can reach some of the agencies". That is certainly true for times counted in hours, and to be honest - for most of the agencies- for a very long time.

Calling everything a "national security crisis" has become a very good career path, and nothing will give your department more funding than national security. An overall perspective and through analysis is what is missing. Press-relases there is plenty of.

Sitting at CDG trying to use the wireless. When the usual first astonishment of how you can build an airport with full wireless coverage but only two power plugs has transformed into just being resigned. I make the following discovery.

traceroute to xx.yy.zz (), 64 hops max, 40 byte packets
1 172.18.19.33 (172.18.19.33) 3.393 ms 2.259 ms 2.171 ms
2 80.10.46.241 (80.10.46.241) 42.893 ms 82.875 ms 42.540 ms
3 10.163.103.199 (10.163.103.199) 42.740 ms 42.346 ms 46.729 ms
4 * * *
5 * * *
6 80.10.46.219 (80.10.46.219) 71.292 ms 44.524 ms 60.045 ms
7 tengige0-15-0-1.auvtr1.Aubervilliers.opentransit.net (193.251.241.181) 45.053 ms 44.469 ms 44.227 ms
8 gi14-0-0.auvbb1.Aubervilliers.opentransit.net (193.251.241.182) 59.519 ms 47.044 ms 66.664 ms

What can I say? What is the 3 hop? And why?

- kurtis -

I am reading an article in the Swedish IDG publication, Computer Sweden on the EU statement on IPv6 addresses. It's a good article with good comments. Until I read the small fact file to the right. "IPv6 addresses will last for 100s of years according to IIS" (the .SE registry). Will they? I don't know. But I do know that unfounded and hyped statements about the greatness of IPv6 addresses and what IPv6 will bring us is most likely one of the blocking factors of getting IPv6 deployed. As my friend Gaurab said, "it's 96 more bits, no magic". People tend to claim there is IPv6 address per mammal on earth and other nonsense. In reality only a /3 is allocated as Global unicast space. And the last 64 bits are used for EUI64. So in reality we have 2^61 bits. Nothing more. There are also other ideas of what to used the unassigned bits in the IANA pool for. Like RFIDs, HITs and so on. So personally I would be very careful with making predictions....