These are unedited transcripts and may contain errors.
EIX ?? 26 September 2012 ?? 9:00 a.m.:
CHAIR: Good morning everybody. I hope we all enjoyed the AMS?IX party last night. Thank you to AMS?IX for that. And tonight there is the NLIX party, so you can feel the same on Thursday morning as you feel about AMS?IX this morning.
Anyway, we have got quite a good session I think today. We have got some business stuff, we have got some technical stuff. We have got some people who are new to the EIX community presenting so hopefully you'll find something of interest. First of all, I'd like to say thanks to the NCC Amanda and MAT who are doing scribing and Jabber scribing, so if you are watching online because we are being broadcast, Matt will relay any questions you have got at the Q and A sessions. So, without further ado ?? if you want to do an IXP presentation at the very end, can you please prepare a one?page slide with contact details, upload it and then come and talk to me and Andy in the break so that we know that you are going to talk so we can call you up on stage.
Anyway, the first thing is mark from NLix who is going to give us an overview of the local peering scene which is no longer just AMS?IX these days. So, without further ado...
MARC GAUW: Okay, good morning everybody. My name is Marc Gauw from NLix and in preparing for this presentation I first issued a very boring draft because you know this is RIPE so I just stick to technology, facts, figures, procedures. But then Andy said well, Mark, this is too boring; you should entertain those folks because they had a party at AMS?IX and they want to hear all about the Netherlands, how good you are, what's going on here, so make it a very arrogant show so that everybody walks away with okay, so that's the Netherlands.
If I'm a little bit too arrogant today, it's not my fault, go to Andy.
Anyhow, it will be an overview, so that's the boring part of the presentation. But what's going on in the Netherlands is very entertaining. Here on this slide, you see all the parts in the Netherlands who play a role and I will take them one by one. But before do that I will introduce to you what the Netherlands is anyhow.
Okay. Well, if you fly around the globe, it will be a hard time to find where the Netherlands is. Because, it's just a very little dot on the map. If you fly above Europe, it's already much better, I hope you can see it. Then the Netherlands is a little bit bigger at least to see it from a plane. And then you see here our flag, and about ten years ago, the Dutch Government decided that the logo would be also used for the Dutch flag so, that's the first update. We have 16.8 million residents and a lot of those folks are on the same square metres. So the whole country is probably just 200 kilometres wide and 300 kilometres long, and all just that size should entertain 16.8 citizens.
You won't believe it, but 55% of the country is below sea level, and I think in this room, you are just about safe compared with the minus?six metres below sea level that Amsterdam is compared with the North Sea.
We have capitols, capitols with an O by the way. The overall capital of the Netherlands which is Amsterdam, which is where you are today, but there is another capital in the Hague, where the Government resides and that's also the Headquarters of NLIX, so some kind of expect. We are doing a pretty good thing in the world. We are just one of the very few countries left in the euro zone with an AAA rating together I believe with Germany, Finland and Austria. We make about 35 K euro per year salaries. I get much less, but the average citizen gets about 35 K per year. And we are on the number 5 ranking of competitive in the world compared with other countries from an economy point of view.
So this is the general stuff about the Netherlands. But at the bottom of the slide you see things which I really am start to go get proud about. We think, and not only me but probably all the IXPs and Internet Exchanges in the Netherlands that we are becoming rapidly the Internet harbour of the world. Well is that arrogant or not? Why do I say that? Well, apart from the facts that the Netherlands citizens are almost 100 percent on broadband Internet, you can state that about 90% of the big international players have some kind of presence in Amsterdam so. If you would like to interconnect to other big global players in the world, if you are in Amsterdam you can probably already cover about 90% of those players.
Above all, we see more and more professional data centres to house all those big players. All the big companies are expanding on a yearly basis. Telecity another building, interaction another big, Equinix another building, a lot of new buildings even in the regions, so probably a lot of companies also need housing in the Netherlands which all needs inter?connections.
And last but not least, we have 2 IXPs in the world top five in the Netherlands, who, together have close to 900 members of which about 700 are unique members. And I'm not sure, but I think that's also a world record.
Still not too arrogant, Andy?
If you then look to who the players are in the Netherlands. I had to restrict to the Internet Exchanges who are listed by PCH. I am not sure whether you know PCH, but PCH is a firm in California, in the United States, who is measuring Internet Exchanges from a member and traffic point of view. And if use their figures, there are six exchanges listed and upper two exchanges are from the same owner. If you look to the map of the Netherlands you see that all those Internet Exchanges at least are connected to somewhere in Amsterdam but they all specialise in a certain area. Just the top guys, take this area and this area. The middle guys take this area. These guys take the north area and AMS?IX obviously takes the larger part of Amsterdam and a big part of the world and NLIX is taking the country and also trying to go international.
If I take them one by one that might be the boring part of the presentation.
GNIX is listed with 50 members. Mainly focusing on the northeast area of the Netherlands. RIX is focusing on the Rotterdam area. By the way I should also teach you a little bit of Dutch. So the northeast area is... and the southwest area is Rotterdam. So, that's more or less to learn you a little bit of Dutch. GNIX cover 7 data centres RIX covering 4 data centres.
NDIX, the east part of the Netherlands, also a little bit international, taking care of the east side of Germany. With 130 on PCH and 21 members and covering an impressive list already of 30 data centres in this area.
The /TPREURB Internet Exchange, that is /TPRAOERB land in the Laywadden area, and they are now ranked already at number 80 with 38 members, and they officially cover 8 data centres but they work together with the other Internet Exchanges in the EU open programme. So on the map you see also some data centres from the other Internet Exchanges with who they work together.
AMS?IX ?? first of all my compliments with the great new website. It looks great. Well, definitely the number 1 in the Netherlands with even more members than I checked with car a last Friday, so 315. 70% of the members international, 30% national. A huge amount of modern switches, 12 data centres in the Netherlands. International partner connections to many places in the world. And since last week presence in Hong Kong but also Bejing.
Last and hopefully not least, NLIX officially are going to PCH, we are now at 306 members, might be about 20 more but okay, 306. We are rapidly picking up on international members, a lot of national members. To be a little bit arrogant, we might be the record holder from a data centre popping point of view with 46 data centres in the Netherlands area. We launched data centres in Belgium in the spring. Today we announce 8 data centres or maybe even more, more parties we have have more data centres are being added, so, it's somewhere close to 60 in the next couple of months with presences all over western Europe.
Last but not least, we are introducing a distributed global peering programme which is my next presentation.
Okay. NLIX is also the sponsor of RIPE 65 so I hope you all subscribe to our party tonight. You can go to NLIX.net dot drink and subscribe. The wi?fi in the room and in the hotel is sponsored by us and today we have a stand and a coffee in the corner is also sponsored by us, so if you would like to be entertained by NLIX there is more than enough opportunity.
So far, this is my overview of the Dutch peering landscape. Any questions before I go ahead with my next presentation?
CHAIR: No questions? Good. Carry on.
MARC GAUW: Okay. The next presentation is just a bonus presentation compared with the other presentation, because I also already mentioned the word distributed global peering.
Long, long time ago, and I might look young but I still already am a little bit old, life was simple. You had a national Internet Exchange and you if you would like to go international you simply took a transit service and that was all. And everybody was happy with that. No complaints. It was working. Life could go on ever after.
But, you know, the speeds increase, so folks are taking care of their money, would like to peer even more than before because you know, transit is more expensive than peering, so why not only peer national, but also try to peer at other exchanges in other parts of the world? It makes sense.
So, a couple of customers took a long haul to other Internet Exchanges in the world to peer even more traffic away than they did ever before. And some Internet Exchanges started to introduce partners to pick up the traffic already in your home country to bring it to that Internet Exchange. And from a peering point of view, a very good thing. From a speed point of view, a very good thing. From a money point of view a very good thing. And that's probably why it's so successful, and why, for instance, AMS?IX has so many partner already, it's an extremely successful programme and that makes sense. But in the meantime, also the applications changed and more and more applications became delay sensitive. Well, you probably make a lot of international phone calls via VoIP and then you probably don't like to say "over" to your counterpart if you talk with him after each sentence. You probably do interactive video over the globe, and it's nice that you sometimes in sync when you talk or move and the other party talks or moves. Maybe yourself or your kids are gaming and then you don't like to be shot by your call of duty enemy already before you make your own shot to the other party. So, are more parties go international? How more the latency is playing a role.
So, in this picture, you see how traffic could, in worse case situations go if you go peering from one country to another country's exchange. Because worst case, your peering partner is in the same country, but unfortunately, not on the same exchange or your peering partner is in another country and then all the traffic needs to go via two /PWHROELS to reach the destination.
Well that's easy to solve, even with all the partner programs or the long hauls that's easy to solve, because just take more partner connections or just take more long hauls to all the Internet Exchanges where you want to be to reduce latency as much as possible. Fine. It works, no problem. Just do it.
But there is another way to do that. Is by taking Internet Exchanges who are in more countries and switch the traffic either directly locally or switch it directly to the right destination country. That's just another way to do the same. Maybe it's better, maybe it's not better, you don't know.
But at least triggered us as NLix is that nobody was doing it yet as far as we know now. So, we figured why not. And we could not really find out why not. So, we got the opportunity to introduce exchanges in more countries to tryout the sense or the nonsense of switching not only in the same country, but also between countries, directly between peers without using blackholes to the traditional home countries. So that's why I scratched away the word IXP home country here because as far as NLix is concerned we are now a home country in six European countries.
Maybe it's sense, maybe it's nonsense, to be honest, we don't know yet. Anyhow, we will give it a try. And hope you like the trial.
Any questions about it?
CHAIR: Any questions for mark? Nothing from the channel? Okay. Well thank you, mark, for two interesting presentations.
(Applause)
CHAIR: Next up is Kurtis. A new and improved version of Kurtis even.
KURT LINDQVIST: Good morning everyone. I am Kurtis Lindqvist, I work for Netnod and I am going to try to do a new RIPE record. I am trying to do 76 slides in 20 minutes.
And I can assure you, I can do it. The question is: Can you follow?
STENOGRAPHER: I hope so.
KURT LINDQVIST: Some you of you might think you have seen this presentation before. You haven't. Because besides going to the AMS?IX party spent of a the night update it go to include something I promised would I was to also have you v6 data. So, what is this all about then? Besides the changes as we do, we also run one of the 13 DNS route servers and we peer quite extensively around the world and I was playing a little bit around the data we have to see how well this peering thing actually works. So, just shortly what this is. It's one of the 13 DNS route servers is operated by us and we have it that more than 40 locations around the world with Anycast. This is all the route servers around the world, they are fairly densely ?? fairly well spread out. All the route servers has a different philosophy at how we deploy them and why we do T some of the route servers are Unicast, some are Anycast. Some of the Anycast at the route servers are doing Anycast only inside operators networks, others do only the exchange points, some do it at both as we do for example. Some have global nodes that are using no export and only propagate prefixes to the closest peer. Other ones propagate this around the world. We do it on a per peer basis rather than a pro location basis. So, everybody does this a bit differently, and diversity is good.
I think already covered is we have most of our nodes are at the exchange points. We have a few inside carrier networks, either tier 1 providers or in a few countries, there is no exchange point we are inside the carry net point as well.
Map obligatory. This is all the locations we are, so we are also fairly spread out.
We peer with around 1400 sessions, this is from this summer so it's a little more nor now with 650 ASN numbers. We also peer with the route servers. In reality we have much more propagation than we have necessarily peering the network. We are also look for more peers, so, please send us an e?mail, please do peer with us. It's for your benefit. We don't get paid or we don't ?? we have no benefit of the peering with you, but you do.
So, I had a looked at the source queries of the DNS queries coming into all these Anycast nodes and both v4 and v6, and looked at where these ?? for each site we looked at these source addresses, mapped them against the RIRs data to see which country they were originating in I know this is a very poor data source for determining where the queries are coming from but they are still interesting enough to see a little bit of what's happening. It's based on a two?hour and 40?minute long snapshot. There is a lot of queries coming in so we can't look at T even this data was quite significant.
Again, it's not a hundred percent valid analysis but it gives you an indication of how good the peering actually works.
So some observations then. It actually does work quite well. You guys are sort of sending traffic to the locus to the closest destination most of the time. For IPv4, there is a lot of RFC 1918 source that is shouldn't appear at all but they do. And we also see exchange points or nodes at exchange points that are seeing traffic very far away while they should be much closer routing instances. We'll come to that. The nodes that have a lot of global transit providers tend to see for global traffic and traffic from global eye balance providers also as expected. Large IXPs with a big diversity tend to see more diverse query sets. Small exchange points that have more national focus tends to see more national traffic.
The news for this time. The IPv6 analysis, it's very different as you will see. Partly because we don't do v6 everywhere we would like to do v6 everywhere but we just don't have v6 peering everywhere, we can't get transit there or we can't get any v6 addresses. It's partly because I think the peering landscape for v6 is very different than v4. Partly you can tell just because we do peering with people it doesn't mean they send us any traffic. For v6 there is a lot of things I cannot explain no matter how hard I try.
Asia, Bejing, the node in Bejing sees the most queried ?? active sources are SC 1918, and the two top ?? the three top networks sending queries not nodes in Bejing, the first two are US based and the third is UK. In fourth place we have China I have no idea what it is. Columbia, Sri Lanka, all Sri Lanka and, maybe as expected, a little bit of RFC 1918 sources. Hong Kong and Google, 61% of the traffic is coming from Google. You'll see for v4 queries Google is by far the largest single source of all queries we are seeing, and we are seeing a lot of big networks from the region as to be expected, this is HQ IX, it's a very large exchange point with a big diversity of peers in the region, so as expected.
H K IX, the same thing, not much to see there. It seems to be fairly equally distributed. We'll notice that Hurricane Electric does show up in quite a few of these lists, also as to be expected, I don't know. But we'll see it as we go along.
Jakarta, mostly Indonesia based traffic. Perfect an a little bit of RFC 1918 space. Karachi, for the rest Pakistan and there is also a little bit of tiny queries coming from Afghanistan. Seems to make sense.
Kuala Lumpar, mostly Malaysian based traffic, also a little bit of China. Nothing strange. V6 only two networks sending us v6 traffic, maybe to be expected, I don't know.
India: Again, a very large set of RFC 1918 space, sending traffic. The rest is all Indian networks H has to be expected because you are not allowed, no one else is allowed to connect, so ?? anyway...
Manilla, again, all Philippine based traffic, a little bit of US and a little bit of Singapore, maybe as expected. Not much to comment on.
Perth: Also, basically all Australian based traffic. A lot of local peers, as expected, nothing strange.
Perth v6: I thought this was interesting. This AS number is the VeriSign J.G. TLD server. That's as far as I know Anycasted but why on earth they are sending 6 percent of the all the v6 queries and why they are only sent in Perth I have no idea, but apparently they must have some node that there locks on sending v6 queries to us. We can also discuss why a G TLD server is sending v6 queries to the root. I have no idea. I think there is something else in the AS number. It's a bit odd.
Singapore, Google again 22% of the queries are very top one. An open DNS. I had a suspicion this is the Google open resolvers because open DNS is also there and we are seeing these two are being reactive in all the data we are seeing.
Taipei: Also all Taiwan ease traffic. Nothing too odd there.
Taipei v6: Also pure Taiwanese based v6 traffic, as to be expected.
Tokyo: At JPnap first, we have two sites in Tokyo. The second largest query source is CHINAnet. China and they should be sending their traffic to Bejing but that could be based on local policies, maybe not so odd. What is a little ?? JPnap v6, all all Asian based v6 traffic. A big Asian exchange, nothing to odd. This is a little bit more strange though. Not the first again, this is the site that we have for both DIXie and JPix, the two address Internet Exchanges. Two big Chinese providers, again this could be a peering decision, what can be explained this this one, this is CNnic they are our transit provider in Bejing but they are not sending their queries there, they are sending them to Japan, I have no idea why.
V6 for the same exchange points. The UK, I bet that's efficient. Otherwise, as to be expected.
Telstraw is also sending the traffic ?? you will notice that tell straw is showing up for v6 in quite a few locations except in Perth in Australia, we'll get there. It's always good to not peer in your home mark because it improves performance for your end users, doesn't it?
Mongolia: A hundred percent Mongolian traffic. Very good. Same for v6, a single network that does v6.
Wellington, all New Zealand based traffic, nothing strange.
Europe: AMS?IX. We are seeing a lot of traffic from all around. The top three ones are US based which might be a little bit strange, if you ask me. Again, open DNS. V6: I have no idea what LK is showing up here for, but okay...
Another, the other site we have in Amsterdam which is inside the tier 1 provider, that's all the downstream customers, I guess they are preferring the upstream or ?? well, it seems to be the down stream customers as far as I can tell.
Ankar, the university network in Turkey, also fairly straightforward.
Brussels: Level 3, why on earth is level 3 sending traffic to Brussels for? Anyway... they are.
We do have ?? Brussels site is a little bit interesting because the upstream there is Bellnet and the university networks of the world seem to follow this peering policy that you always prefer the upstream academic network, this has a little bit weird side effect that all the NRNs of the world are all sending their traffic to Brussels so. You are ?? at least it used to be if you sat in the university of Aukland, you talk to the server in Brussels. Of course that makes sense...
This, I can't explain. This is the node in Brussels. We do peer with v6 with quite a few other networks but it's only a single one of them who actually sends us traffic. I have no idea why. That's the GP who could send this traffic to us in London. But anyway.
DCix, again Google is on the top followed by Deutsche Telecom. Remember that Deutsche Telecom is actually sending their v4 queries to DCix because they'll come back to us.
V6: Again, fairly distributed set of countries, nothing strange there, Hurricane Electric showing up again.
Helsinki. Nothing odd. Finnish and Russian networks, seems to make sense to me.
London: M, it S India, number one. They are also number 1 in mum buy by the way. I can't explain that.
V6: Number 1, Deutsche Telecom. It's really good not to peer in your country, right, isn't it? But hang on for v4 they are sending the queries to Frankfurt but not for v6. Don't know.
My other favourite here is again tell stray because they don't peer in Australia. And Thailand. So if you think it's really strange that this is incumbents believe that it is such a good improvement for the end users not to peer inside their country then you end up talking to the route server in Europe instead.
Sweden: Only Swedish providers, nothing strange.
Milano, mostly Italian networks, nothing strange.
Milano v6: Again, Italian networks, all local networks and Hurricane Electric from the US. Nothing unusual.
Oslo much: V4, nothing strange, Norway and Sweden and the next slide is the strangest slide that I really can't explain at all.
This is Oslo v6: China, China, US ?? well maybe, I don't know ?? China, JAnet. We peer with you in London. I have no idea why the Chinese v6 networks are sending their queries to Oslo. I have nothing against the Norwegians, it's a really nice place, but you know there are closer place to say home. An JAnet, I have no idea.
Anyway, Paris: Again, mostly French networks, and Swiss, sort of to be expected. Same for v6. I don't quite know this Gand AS, I have no idea what this is, but they are sending a lot of queries.
Bucharest: All Romanian traffic, more or less and from the neighbouring countries. Makes sense.
V6: Also the same, it makes sense.
Sweden: This is also announced the covering prefix Fr. We tend to see a little bit more diverse traffic, which is to be expected, nothing strange for v4
For v6 again, we are seeing Swedish networks, nothing too odd there.
Estonia, Talin, at the RT IX, again all Estonian traffic. V6 the same. Nothing strange at all.
Ukraine: Nothing to to see, more or less all Ukranian traffic, and the same for v6, it looks very good.
Vienna: The same, nothing here to see. V6, looks really good, all neighbouring countries, you are all doing well.
Armenia: Nice.
Middle East: Again, Bahrain we are seeing all traffic from Bahrain, looks good.
Same for Qatar.
US: Looks good, Chicago start?up, all the university networks. The same with Nota, all the sub?American networks. Google, 47%, again they are by far the most active network we are seeing v4 queries. Have you noticed we haven't seen them for v6 in a single slide yet? I don't know why that is.
And Palo Alto in the US, Bulgaria. I don't know... anyway...
DC, this is inside the tier 1 carriers, this is all the down stream customers.
And last: Africa, Johannesberg, all South African and neighbouring countries, nothing strange. The same for v6.
And last South America, all Brazilian traffic, more or less, all to be expected and that's the last slide.
In general all of these peerings do seem to work. I highlighted some strange cases but if you consider how many peerings you have you are doing a fairly good job out there, it does seem to work but the v6 things does some really strange things and I think the v6 peerings and v6 routing must still be a little bit broken.
That's with we see. That's how it works so you guys are doing a good job. If you want to do a better job, then please do peer with us.
Any questions? No? Someone?
AUDIENCE SPEAKER: Hi Kurtis, this is mike from Google. So, I'm not an expert on this at all so maybe you can help me. Do DNS resolvers use latency to try and determine the best place to pick ??
KURT LINDQVIST: They are supposed to I think is the answer. It depends on the software as far as I know. It very much depends on the software and versions but they are supposed to but they don't. That's the simple answer. It's a little bit more ?? it's a little bit more complex than that, but it all depends on the software version. I think it's getting better with more modern software and they are doing it more and more.
AUDIENCE SPEAKER: So some these issues could be not routine, they could be DNS servers picking servers that are further away than the closer one.
KURT LINDQVIST: Yes but the latency is also tied to you optimise your routing. They do go hand in hand.
AUDIENCE SPEAKER: I think it's a comment I have said before, but the country of the AS registration isn't necessarily the country where the queries originate from.
KURT LINDQVIST: Correct. That's especially true for Google.
AUDIENCE SPEAKER: People like level 3 and so on have big European networks, the fact this they have stuff appearing in Brussels isn't perhaps a big ??
KURT LINDQVIST: That's right. But we are not seeing any level 3 anywhere else. Everything I think sell query from an AS is going it Brussels. I agree.
AUDIENCE SPEAKER: I am curious, you mention that you said in a plane private address space for v4. I did not see any strange solves for v6 traffic.
KURT LINDQVIST: I have to admit because I wrote these slides last night I didn't have time to look at that. One thing I wanted to pick out how much is 6to4, it is quite a bit of traffic is 6to4.
AUDIENCE SPEAKER: Because once I looked and strange source and I noticed plenty of documentation prefix and surprisingly unclear link up ?? so it would be interesting to compare with your data.
KURT LINDQVIST: I can go do that. They can't be that much because it would have showed up on the lists here. So, I mean it's not as much as as it was for v4 because it wasn't one ?? what I did was I looked at anything that was, the threshold was 1% of the queries and none of the they shall hold was strange traffic, but it's a good thing. I should go look at it as well. I am sure there is some.
AUDIENCE SPEAKER: Patrik Falstrom. Let me expand a little bit and answer on the DNS question. Just because we do have extremely homogenius both data and software and hardware on each one of those nodes, we do seriously believe that the amount of application layer soft are back I don't have is actually pretty low (off) but on the other hand, you have absolutely right that you have to be careful when drawing these kinds of conclusions on why someone is ?? why A is communicating with B. But as we are just looking at the IP traffic between the AS, its selves we think that's fair. I am doing a similar analysis in parallel regarding M, it P traffic and there I am having these kind of issues that you are talking about that people are sort of ?? I don't really know yet why a certain choice of what server to talk to, but for DNS we are pretty sure that this is good enough data.
CHAIR: Okay. Thank you Kurtis. That was 20 minutes including questions. So I think we do have our record.
(Applause)
CHAIR: Next up is Miriam from the NCC Atlas project and she is going to talk about traffic patterns during the recent Olympics.
MIRJAM KUHNE: I am from the RIPE NCC I am mostly doing everything that's related to RIPE Labs so all the statistics, prototypes, tools that we are doing we put that on the RIPE Labs website and that one was also one interesting project, that we have done together with U RIX and how many of you have been at the EPF in Malta last week? At least not everyone so that's good. Some of you this will be new because this was presented at the EPF last week I believe it was or two weeks ago.
What what did we do? Just to give you some idea how this started, because, Serge was U RIX he started looking at some traffic patterns at the exchanges during the world cup in 2010, and we continued that basically this year during the European championship and during the Olympics. So, at the at 2010, Serge when he was at U RIX he basically looking at all the exchange, all the traffic patterns at 51 exchange points, members of U risks and he published a long report of that and he also referred to that on the RIPE Labs website. And he found a few interesting patterns.
For instance, this one where you can already see a little double dip that we also later identified again in the European Championship this year. So that's ?? it was just a different pattern than a normal day, so that was interesting to see. Then the other thing you noticed was at the time, a huge increase, for instance, at the lone Nap exchange in London during one of the matches.
So we took that as inspiration basically and sat together with Bijal from U RIX and a few of the exchange points and just before the European Championships started and thought well what can we do, how can we make some interesting measurements? And Bijal reached out to all the exchange points and asked for participation, so 15 exchange points agreed to give us the data. So that we measured, we basically collected the data on a daily basis and then looked at certain patterns during the Championship, during the matches.
So during the ?? the European Championship, the football was different than the Olympics because the football was usually in the evenings, there was some clear time scheduling for the matches and so we saw some interesting patterns there. Whereas the Olympics, it was more diverse, I'll get back to that. And so we published five articles during the football Championship earlier this year, and looked at, you know, the stages that are listed there on the slide.
And what we basically found is a lot of exchange points that we looked at, we found this pattern of either double dip like on this slide where it looks like people just got off the Internet and did something else, for instance, watched television during the football. And then a little dip again during the break, where people might have gotten back on on the Internet and then another dip during the second half.
The opposite pattern that we saw also was like a double bum, sometimes where you could see people watched the entire match on the Internet. And this one is the actually ?? or the one before, I didn't mention it, it was Portugal, that was actually during the Portugal and Netherlands game, so that was quite a popular game to watch apparently. People didn't do anything else on the Internet during that time, it seems.
This one was in AMS?IX and even there, even though there was huge traffic volumes there, you could still see a different patterns. We looked at, like the normal traffic patterns at the exchange points and then we looked for different, basically different patterns and the little green line in the back was a week before the game, just kind of identified it as normal pattern during a day like that and then the red line was on the day where these two matches happened. And you can see quite a different pattern there.
So the other extreme. This was kind of our favourite exchange and some of your exchange in Russia because they always show huge peaks during the matches and you can see there, like, it's double bump that we identified during one of the matches, which was N this case ?? yeah, it wasn't ?? there just seemed to be independent of the actual matches. People just seemed to really watch all the matches there and we saw a big increase in traffic in that exchange point.
The same here. That was actually Ireland, that was during the Ireland?Italy game, we also saw a huge bump there, a huge increase and then during half time, you know it it kind of went down and people did something else or ?? I mean that's just speculation, we just looked at the traffic patterns and what that actually means and what people really did is a different story. It's interesting, after we published this article, a number of people added comments from their perspective in their countries and added some additional information from their local perspective how the situation is in their countries and that was really helpful.
That was just ?? that's the last thing I wanted to show. There was one ?? we can't really see it. There was one with penalties basically where we could also see during the first half, you know, and then you see the first half, the breaks, the second half and then the penalties. It's basically a very different pattern than other days when in other games it didn't happen.
Okay, so that was during the Championship. I wanted to show you some of these graphs.
We looked at many others, but I just picked out some of the more prominent ones.
Then for the Olympics, you are thinking how can we do this because there are no specific times, during the football, there were like two matches everyday, but the Olympics just went on during the day and at various times and people are at the office, and so, how can we identify different patterns there?
So, there was a bit of a challenge but what we basically did is we collected again the data of all these IXPs that participated in this project and then we tried to match it with the games that happened during this day and kind of tried to identify what were people interested in or can we find something in the Olympics schedule that kind of matches the pattern on the exchange point and the traffic graphs that we saw.
And so we looked at, again, looked at the typical patterns and a few weeks before and a few weeks after and then that particular day, and checked if there were any difference in traffic to see. Like, for instance N this one, this is lone app and you can see on the right?hand side there a little red graph, that's during the open ceremony, which was much lower than normal traffic than the traffic on other days at the L O N AP exchange, which was interesting. The other thing that you can see there on the same graph, the same day a week later was actually during weekdays, you saw a huge increase in traffic at the lone app exchange ?? during office, during the weekday during office hours and we were curious to find out what happened on that day, what were people maybe interested in seeing, and there were a number of games happening during the day that people apparently wanted to watch. And because it was during office hours, they might have watched it online, so, that might have created that increase.
A few other patterns we saw like for instance in INEX Ireland, apparently you are interested in boxes, they were just two boxing games, one week and then also the week after where we saw a peak in the traffic during one of the boxing games.
STENOGRAPHER: Go on Katie.
And also what we could add from the RIPE NCC's perspective, this is basically IXPs giving us their traffic and we kind of looked at the traffic patterns there. What we wanted to add from the RIPE NCC was the RIPE Atlas project that has been mentioned a few times before during this week also. So we have many probes sitting at various places in the network and so what we did in this case, we looked at all the 161 probes that are located in the UK, and we were curious to see which the traceroutes go through the IXPs and how many go through other connections, because obviously not all traffic goes through the IXPs, they are also private interconnect and other peerings.
So, about half of the probes took other routes, but 28 of the probes went through lone app and 55 went through the links. So that's kind of an interesting data point to see how many trace routes, trace routes that we asked the probes to do actually went through the exchange points. That's kind of an idea what kind of data we're looking at there
That was basically it, I wanted to show you was really a fun project to do. Thanks to all the exchange points that participated and also thanks to those who added their, you know, comments after we published the data and the speculations that we had about some of the patterns. It seems to really show that if ?? I mean, you have seen this before, I mean also weather has an impact apparently and some other events, you know, really has an impact on Internet traffic. And so you can identify some patterns during major events. But overall, it doesn't kind of break the Internet or something, it seems to be everybody can cope with that despite like big peaks at times. So, yeah, it was a nice exercise to maybe also identify these kind of traffic patterns for other events in the future. Thanks to all the exchange points that participated. And last but not least, there are some links, maybe the last one here, that might be interesting to point out because we also published all for the Olympics, not for the football but for the Olympics we published all the graphs per day, per exchange point and you can basically see a nice overview of all the exchange there on that page for those times during the Olympics.
And that was it.
CHAIR: Thank you, Mirjam. Any questions for Mirjam?
AUDIENCE SPEAKER: Thank you Mirjam for a very interesting presentation. I think a lot of the traffic is probably kind of hidden because it was delivered from CDNs that are inside access provider networks. For example, BBC were using a variety of CDNs to deliver their streaming stuff. It would be really interesting if there are any access providers in this room who would be willing to give presentations about the traffic impact that they saw from major sporting events like this, because to a certain extent, you can't really see the traffic unless you are looking at it from the access provider perspective, and I suspect they saw quite a lot more of a peak than we are seeing more publicly on the Internet Exchanges and so on.
MIRJAM KUHNE: That's true, we realised that, that's why I said not all the traffic went through the exchange points. But it gave and idea what happened.
AUDIENCE SPEAKER: From Comcast. We didn't see a huge change because of time of day shift.
CHAIR: Of course Red is an American access provider rather than European, so it makes a difference. We have got UKNOF in a couple of weeks time in London, there is a couple presentations from CDNs, so, that will be available on webcast at the time and later if people want to follow up on that. I think, you are talking there as well, aren't you, Mirjam?
MIRJAM KUHNE: I'm probably going to show something similar.
Thank you.
(Applause)
CHAIR: Next up is Maksym from AMS?IX who is going to talk about data around IPv6 day.
MAKSYM TULYUK: Hello, my name is Maksym. Before I start presentation, thank you that you made a huge effort and come to listen to our presentations after AMS?IX party yesterday. So, I will talk about what we see during and after IPv6 Day before I carry on, does anybody don't know what IPv6 Day was about? Okay...
So, how we perform. Before IPv6 Day I used this snapshot because it looks much better, much more impressive than ??
AUDIENCE SPEAKER: Clarification question. Do you mean IPv6 day or IPv6 launch?
MAKSYM TULYUK: IPv6 launch day. So it's what you see ?? it's what we see before IPv6 launch day, and that's with wee see during IPv6 day. So, as you clearly see on this graph, it was 50% more traffic. And when you see ?? when we saw it there was a simple question: Is damper on this day and carry on for a long time? It stayed for a day, for a week, and a few weeks more, so, it's really easy to make a simple conclusion, that yes, it's come. This 50% more comes and stays for a long, long time.
And then if you look ?? and we look for three months snapshot, you will see that firstly, it's after June it went down and in July it started going up again, and when I first looked at this graph, I saw, okay, it looks something similar, what what's it about? It looks like IPv4 traffic. So I put ?? the two graphs on the top is the picks graph, below is IPv4 graph from our exchange, and you see that traffic is followed quite a similar pattern and it makes a simple second conclusion, that IPv6 traffic, what we start seeing after IPv6 day is more traffic that carries some real traffic. Because if you look, at traffic that was before June, it was all the time same line, with a few drops.
However, if you look at the picture in general, all this big graph is IPv4 traffic and only in a bottom there, we start seeing IPv6 traffic, so for now it's 0.5%.
Short summary: 3 points. It's first that traffic come and stay after IPv6 day.
Second, it's now became quite similar to what we see in IPv4 pattern. And it just 0.5%, but it shows some way to grow.
Any questions?
CHAIR: No questions? Thank you Maksym.
(Applause)
CHAIR: Our last item before coffee is Cara who is going to talk about some business process stuff and the AMS?IX collaborative model and... just a reminder for people who have come in since the beginning, if you are from an IXP and you want to talk in the final slot, can you up load a single slide to the presentation archive with some contact information and come and see Andy and myself in the break so that we know you want to talk. So...
CARA MASCINI: I guess I'll start with ?? so, I hope you had a good time last night. The AMS?IX management team had a good time, so we hope to see a lot of those pictures ?? they will be on Facebook on a special page that we'll send to everyone, so, thanks for being there and having fun.
So, my name is Cara Mascini, for those who don't know me I am the Chief Marketing Officer at AMS?IX and this morning I am going to talk to you about the industry collaborations.
Now, when Fergus and I got talking about the topic I'm like I'm not sure that this is the right audience for that sort of topic but he convinced me that I should actually talk about it. What I'm going to use is some of my experience in the last two years. I have been doing an MBA and a lot of that is reflected ?? well, some of that is reflected in this presentation and I wanted to share some observations that I made during those studies that are very, very relevant to our industry, so, therefore, this presentation.
These days in our business, in any business, we see more and more uncertainty. There is financial uncertainty, there is commercial and market uncertainty, there is a technological uncertainty even. There is a high pace of change. Basically what we see today might be different tomorrow. It might even be different this afternoon. Someone could have launched a new service, it's going to be very ?? it's taken up very fast and it's the next big thing. And yesterday it wasn't even there, it may not be there tomorrow even again.
So, in our industry, especially in our industry, and I don't mean maybe the Internet Exchange business but the Internet business, e?commerce, applications, all of that, there is hypercompetition, there is exponential evolution of applications. And there is very, very little time to get settled. There is also an increased transparency, everybody knows everything. Knowledge is more or less a commodity. It's not so much about the knowledge that you have today. It's more about the knowledge that you can create for tomorrow; the creative solutions that you are going to be able to develop.
That's the world we live in today.
So a few of ?? a few gurus, there is many of them, but I took a couple that rang true with me. And because they have such a link with the industry that we're in, that the Internet Exchanges, but also all of the exchange members, the data centres, the ISPs carriers, everybody in the business.
Now, there is three people that I chose to address here and that's Josephine Greene. Josephine Greene is a VP at Phillips here in Amsterdam. She is head of strategy for the design ?? well, branch of Phillips, and she started this movement that is called Pyramids to Pancakes. I don't necessarily like the name, but the idea that is behind it is very, very interesting. What she talks about it to build flats, very flat social structures, open ecosystems of collaboration. So, not large Pyramids, you will, of organisations that have, you know, the mingle is profit maximization to build a sort of pyramid around it have people in many many levels. No, what she advocates is an ecosystem of collaboration where, that is concerned with access, not possession, and about sharing knowledge and resources and this movement is called Pyramids to Pancakes. It's very interesting, if you just Google, it you'll find it and it does ring a bell. Also in our industry where we see a lot of different parties working together, smaller, larger, everybody is creating this ecosystem together. It's open. It's social and it's very flat in terms of hierarchy.
Now the next one is Gary Hammell. Probably many people know him. He is a Professor at MIT and he has been quoted to be the most influential business thinker of today by the New York Times, so I'm sure some people have heard of him. But what he has been doing for the last couple of months is very interesting. His ?? his thinking is basically that you have to build organisations that are fit for the future, but not just fit for business, but fit for humans. Humans are adaptable, flexible, resilient and inventive and companies should be places where humans would like to be and can actually express their abilities, their createivity and so on. And what he says is that we need to ?? well, companies need to reinvent management, management in general, and what his observations are that the web, in fact, has all of the features that management should have, and I'll go into that in a little bit more depth, but he says that the web makes us all social and work together, it makes us accountable because it gives transparency to everybody. It's a very interesting thought.
And then the third one that I chose, it Muhammad Yunus, some of you may also know him. He is a Nobel Peace Prize winner in 2006, he is the founder of the Grameen bank of microfinance. And the Grameen bank actually gives power to the people in a sense where they give poor women mostly access to finance to, you know, grow crops, buy cows, whatever they need to do, they are a small business and to grow their small business larger.
He says ?? he has an idea of forming social businesses and social business, it's a business just like any other business, but it doesn't have a profit goal. It has a social goal. And the social goal could be, for instance, you know, solve poverty, which is the Grameen bank's thinking, but there might be many, many other smaller goals as well that could serve this purpose. And this also rang a bell with, you know, the industry we're in, the exchanges in my case, where we in fact also have a sort of a social goal, if you will, and we are not non?profit owned and our goal is to basically make sure that IP packet inter?connection is facilitated for everybody all the time. And that is in itself is a sort of a social goal, if you will.
So those three people, even though their perspective are very different, their business model invasions have a number of things in them that come back in each and everyone.
And I have taken four of those that are also very true in our environment. One of them is open. By being open and fair, you give access to everybody, you share your knowledge. You create a transparent level playing field for everybody.
And fairness brings faith and fairness brings trust and trust is the result of transparency, accountability. So, by creating that level playing field, you create an open atmosphere for everybody to prosper in.
Being flexible also, you know, the ability to be able to accept change. In this industry, cannot do otherwise. You have to accept change. You have to be able to adapt to be resilient, like people are resilient. These business model innovations tall take flexibility integral in the business model. And I think that's something that we have learned in this industry that that key to be successful.
Learning, of course learning from each other, being able to integrate knowledge from others and to apply it to your own situation, that's a very important one as well.
Third one is of course, almost by default, innovative. In these days, you have to be innovative to be successful. You know, knowledge that you create today is gone tomorrow, or new knowledge might have been created by someone else that you can use to improve your own business or to bring you further words to your goal.
Also, and that's interesting, is that you have to be somewhat contrarian, so challenge the dogma, if you will. You know, the innovative solutions almost never come from the status quo, the people that are the large players in the market. Almost always, they come from the fringe, and you have to challenge the status quo to be able to get further.
And the fourth last one, but not least, is social. Well, I guess that we're as Internet Exchanges, we are very well known what it is to develop a community. And inter?ependent sets of members, customers, participants, whatever you want to call them, who are able to collaborate together and bring, and stimulate and increase successful collaborations.
So those four: Open, flexibility, innovative and social are the main points I took from those three business model innovations, one being Pyramids to Pancakes from Josephine Greene, the other being reinvent management from Gary Hammell and social business from Muhammad Yunus.
Now, going to our industry. We have many, many different types of parties in our industry. Some commercial, some non?profit. Some there to provide end user services, some to provide wholesale or very technological core services. There are vendors that do new developments and research. There are registries, NICs, there are a whole bunch of different types of organisations but we all work together quit closely and especially these conferences are when things start to happen and where the new innovations tart to bubble up. And we all need to get each other to make the Internet work, right? So it's almost inherent in the companies.
So you see, a lot of industry collaborations. The exchanges themselves are of course a perfect example of it. Most of them founded by a bunch of national ISPs that were competing with one another or maybe some other parties. But there is so many alliances, forums, associations and whatever in the industry, whether they be mobile like the GSMA or infrastructure related, like the ethernet alliances, or whether they are more of a platform of discussion, yeah, there is a whole bunch of them.
And then there is the different collaborations between the parties themselves directly. There are carriers working with carriers, exchanges working with exchanges but also exchanging working with carriers, data centres working with exchanges, well there is a whole set of different collaborations out there that our industry sees.
Now coming to the AMS?IX /THRAB tif model, /PWAU that was (collaborative) one of the topics that Fergus wanted me he to address.
I I have also shown this last week in Malta. This is the, what we call the AMS?IX collaborative model where we ?? well you can see, we work with the Amsterdam based data centres, we're in 12 different ones right now. Basically we have our set of members and customers, those are in the lower bottom. And then we have a whole different array of type of /KHROBrations that we do (collaborations) this that we do with, they could be either carriers, transport providers, ISPs, data centres, and they all have, well we have different names, a lot of them are linked to what we call the partner and resell err programme. But we do, for instance, something that's called AMS?IX enabled data centres, that's where we work together with resellers and the data centre that are not necessarily very close to Amsterdam, but that enable parties to indeed come to Amsterdam easily.
And then of course we also have the new exchanges that we formed, especially for instance AMS?IX Hong Kong, AMS?IX Hong Kong is a collaboration between ourselves and a carrier. It's a hybrid model. It's a new type of model for an exchange. We are experiencing how it's working. We are experiencing how things are ?? the new things that we have to, yeah, basically learn and then we can use those things that we learned in our collaboration /W?S other parties. Like the just announced resell err ship with C come for instance in east Africa.
So, all in all, it's a model that evolved over the years. We started with this partner programme back in 2004, and we slowly learned what works, what doesn't work, technology also helped us in applying new solutions in that area, and all of that came up into that model.
Now, what I did want to note is what we believe in and that's beneath the model, if you you will, or above, whatever ?? in fact it's inside of the organisation and hopefully transpires to all of the partners that we work with. We believe in the neutral IX model, the independent not for profit model. We collaborate with others and others could be anyone in the industry that has an idea that works and that fits with the ?? well what we believe in, basically.
So markets need development or stimulation, we are open to that as well. We have helped many many other exchanges in other locations in the world so set up and just tell what we think works. And then there is ?? and then what we like to do ourselves is build principal market places like we have experience with Amsterdam in other places in the world. Benefiting the whole ecosystem and not just a few, so not just one or ?? but one carrier, one data centre, or one exchange, but a whole ecosystem of everybody in that particular market.
That was it.
CHAIR: Thank you car a. Any questions? Not even any trouble makers in the back row? No? Okay. Well thank you again, car a.
(Applause)
CHAIR: So, break a couple of minutes early for coffee. If we could be back for 11 o'clock for the second session. Thank you.
Coffee break.
