Tuesday-Amal-2PM
Tuesday, 28 October 2008 14:00
EIX WG
**
The EIX session commenced after lunch, as�follows:
SPEAKER: OK, everybody, I think if serve ready, we are ready to start. The EIX Working Group. Somebody pointed out the information appearing on the screen here was wrong, and it said this was DNS. DNS is next door in the other room and this is the EIX Working Group.
So, welcome to everybody. We are spread over two sessions so we are here now for an hour and a half and then we are here again in the morning for another hour and a half session. What we will try and do is cover the items A through F if we can,ed to, and then the remainder tomorrow. Has anybody got any other comments or anything else to add to the agenda? There is only one difference in the IXP updates and that is /PH*EUBG isn't going to give because they have more detailed stuff tomorrow and that is certainly what they said to me earlier, I believe that is still the case. And they will be replaced instead by NaMeX, one of the other Italian exchanges, give an update on what they are doing instead. Otherwise, has anybody got any other comments or anything else, anything else to raise regarding the agenda?
If not, there is one other add� bit of administrative trivia, this is the last time I will chair one of these sessions because I will stand down as co chair of EIX at the end of this meeting and Fergus Mr.�/HR* take things forward for RIPE 58 and find somebody else to replace me next May so this will be the last time I am stood here.
I think without further ado, Jabber describe is there and that is Roomy and France from the NCC is our meeting describe and of course we have got the wonderful Stenograph /AOERBGS, and without further ado I think we will continue do and ask /TPWAUBG the role of IX in the development of the Internet in corps and then we will have the IXP updates. Because there is so few I think we can probably let people have more than two minutes. We were talking to somebody yesterday from, Google and what we worked out was that he was reaching 0.8 Curtises on the scale he was presenting.
Curtis: (K) thanks. So, a presentation I did earlier this year was about the how internet exchange in Europe came to be and change of traffic came to be in Europe and I did this originally to see if there was any lessons that could be learned for other parts of the world and other regions and why this came to be at all so it is more history into this, and this presentation is� might seem like a statement but it's kind of a a question to see if people agree both in my view of history and the conclusions but we will see about that and I will try to speak a little bit slower.
First of all, so this presentation came from talking to some old time people, people older than me in the industry, not� maybe age wise too, but tried to do some observation about what actually happened when peering came to be in Europe and when they the exchange point was first established and tried to see what can be applied in other parts of the world and other parts of the world are similar to currently in the development phase where Europe was earlier when this started to happen. We all know interconnect between operators is something between transit links public peering and private peering and there is of course reasons why all of these are used and when one over the other.
When I talk to people there seems to be some sort of agreement that the exchange of traffic in Europe came in three phases and we had early and most academic days, there seems to be some debate and vary a bit between country to country, the first 92 to 94 and 93 to 95 and their only commercial days were basically up until the late 90s and we have modern times where we are still today
Q. In the early phase of academic days basically no real competition, people wired up and connected wherever possible and there wasn't actually that many connections to start with. The people who want to build out IP net was all corporated, they all learned each other and shared pretty openly and they also had to connect was the only way. Traffic in these days mostly mail and use and use distribute of use, people gave me colourful stories about the equipment of the times and they used very colourful wording about the equipment. Basically, yes, people corporated was very open.
I don't know if you can see the slide very well but this is� network diagram of one of the first interconnect points in Amsterdam where at the bottom you have some surf net and EU net, you have some E bone networks and all these people connecting and someone told me allegedly, still bolted to the wall, I am not sure if that is true or not. Maybe Rob can confirm. This was the first start of the combined� traffic exchange in Amsterdam a lot of these started in Amsterdam and historical accounts told to me it was simply only place where people had enough equipment to change traffic, only place where physically this could be done. A lot of links were slow speed links out to the other countries, going from Amsterdam to other parts of Europe.
In the early commercial days, and here also, the store see seems to shift a bit, some countries we still had the strong university network academic network that had all the funding for the network build, there was basically some sort of national subsidies to run this. In other countries, the funding shifted from going to the university network to the universities themselves and they could choose to either buy services from the national research network or from any other commercial providers, which gave the first steps into market of this. And we also saw some of the first peering policies, networks difficult to peer with each other. We also saw some rules being created that people would only peer and exchange traffic in it makes sense for both parties, win win situation of mutual benefit. The first started to make a difference between who they peered with and they also started to crystallising their own service. I was told a story that EU net, they sues use this had a selling point birth in the UK had D /TK FPL E net had trance at LAN I can from Amsterdam and from the UK, didn't want to use the commercial offerings of each other so refused to peer, there were a lot of personal stories here but you see the first emerging service offerings that I saw in the peering.
We also had the first depeering threat, this is from March 1995 it's a newsletter from Dante where they� the line at the end "unfortunately Dan at that cannot indefinitely free and unlimited connectivity to some networks while charging others." This was eventually they realise that had they had mixed up transit and peering and the e bones, or Dante side realised that if they did depeer EU net they would both said send the traffic to US and incentive setting up private peering increased and it remained one of the longest running private peers in Europe, there is something to take from this. I would still see these happening to day but a lot emerged from these early days. (D A N T E)
It first of all it was a way to save costs. The first cost was the transport capacity, I was told stories as late as today and also earlier that it was cheaper to buy connectivity from any European countries to the US than a circuit between two of the incumbents in Europe so the transport pricing itself drove people to interconnect where they could in countries and also to try to lower the traffic volumes. The other part to this was all the Europence of course had to pay the big US providers and by slowly as the Telecom market in Europe had deregulated could you afford to this interconnect in Europe and some people persisted in doing that even though costs quite high in Europe and it just in order to lower the transit prices they had to pay to the US. And eventually, some of the European networks got in equal terms with the US networks and got peering in the US as well but that was much later.
In these days, also, in international circuit with fairly low band width so peering in itself had no bearing on the lowering of RTs, that is highly simplified point but roughly.
Right. We also another shift that happened more towards the modern days late '90s, early 2000, originally when I joined this industry there was a rule 80 percent of all traffic emerged in Europe went to the US and only 20 percent stayed in Europe, during the early dotcom boom years and later a lot more was added in Europe which also led people to stay more locally or at least inside their own language and culture barriers and this availability of local content and the need to exchange traffic inside a country, of course also� this is my guess and my estimate� probably helped change the interconnect landscape and also this was possible because Europe had this exchange points and they had grown out in most countries and could you reach the local content over fairly low latency and fairly easily and this again was a good circle as opposed to bad circle and it helped feed more services being developed and you see a lot more these services become available and the language.
This is the following two slides are made in 2001 and I honestly can't really remember what the number stands for any more but the second number, I found in my notes is actually, when I use to work in� no longer exists but we had six fibre rings in Europe and the second number is the percentage of the traffic that didn't� there was generated as a customer in that country, I think the first number is the traffic that stayed object that ring and basically a Scandinavian, central European towards Spain and so on. And the difference of traffic that made it to the US. But it's� it's clear here that in 2001 already the 80 percent rule no longer applies. You seat traffic is starting to stay inside language and cultural regions. Interesting example is Finland where my guess at the time was because the Finns don't actually� I am Finnish I am allowed to say this� English language is understanding is less at this time than perhaps now and might explain why number of traffic staying in the country is very high, it's also the case Finland was very early with having local language services, so I am not saying this in a negative way, I think it helped contribute. There were a lot, lot more local language content available.
AUDIENCE SPEAKER: This is mike here. Where is Denmark and what has happened to the UK on that map of Europe?
SPEAKER: Denmark was actually originally made in PowerPoint in really weird colours and I couldn't figure out a way to import it so this is a grab screen for the� for some reason those colours don't stay. It's nothing against UK or Denmark, they are just a bit special. I could make a joke� never mind.
This again slide from, I took data point here a lot of this is, this content of course already helped contribute to this. It became more important for the pride providers keep traffic locally and we saw onset of� difference updater routing where people started making these differentials.
So my claim here, and please shoot me down afterwards if you don't agree, it's hard to prove that these dense interconnects and these emergence, because of the highly regulated Telecom market force people to do peering, actually helped people and helped Europe grow quite a bit and it also having these local connect it also helped the European providers to create an own edge over US base providers and even when I was working in quest we still had US providers who came to Europe and just because they had US traffic assumed they could get peering in Europe which wasn't as easy as they thought for European customer having cheap and fast access at the time to European was a selling point. Today that has changed a bit but at the time it wasn't as easy and I would like to make the claim that the development in Europe it was not mirrored in the US and it was unique to Europe and do I think it is more applicable to what we are seeing in other regions of the world today, more similar to Europe at the time than US because US never really had any of this happening.
Again from peering well you do get this keeping traffic regional, you might create incentives to local content and object the other hand more local content become available you will benefit from having local peering. There is of course a redundancy claim saying exchange point in the country is good for national strategy when it comes to redundance and robustness.
Peering, why? You can do it private or public but the establishment of the exchange point in Europe, 36 in Europe,
AUDIENCE SPEAKER: 105.
SPEAKER: In Europe?
AUDIENCE SPEAKER: Yes.
SPEAKER: Anyway. It's even better point this. Density of exchange points is is a lot more than you see in other regions of the world but a lot better connectivity than you do, and it's also help grow the establishing of neutral ground where carriers can go in and meet and do this traffic exchange and also therefore having easy access to other services that of benefit like local content that go:go into the same neutral grounds or other goods like TLD services etc.. I am going to skip some of this stuff.
So does it actually make the difference. The following numbers given to me, it's small Asian provider, who connected to Linx in London and without any or little effort 11,000 routes directly from the route serve he and managed to get 40,000 routes, people with completely open peering policies and they basically had no traffic to offer so they couldn't really go and ask for that as a reason but they still got 40,000 routes by just connecting which to them was a lot and to them made a lot of sense to do peering although it was quite far away and cost quite a bit. Peering doesn't always make sense to go far away but it can� it's worth doing in numbers and looking at it. What do you gain from doing this traffic in region or even out of region.
Peering naturally basically always make sense except if you are come /TKAPBT principal provider or incumbent. If you are incumbent and really worried and really think your regulator should help you by stopping this silliness of exchanging traffic for free, if you are dominant provider who is acting in a regulated market I have some news for you: Your customer share can only go down. If you start with 100 percent and� it it can never increase. You can of course do anything you can, most incumbents have this thing they do to try to stop this as far as they can, it won't help the local region or network or the customer base there. So, I am sorry but I am not really crying for you.
Others might.
Regulation, government tended to, even in Europe, as well, when this happened, they can� kind of like to regulate, that is why they exist, when it comes to this traffic exchange and for free, it's very little they have to regulate, very little interference with the regulate they are a makes a difference. In Europe he had a list of Swedish implementation, Telecom director that says that� whatever it's called� Telecom director� basically stipulates that a provider cannot deny any other provider an interconnect but they are free to price it as they want which I think is actually good enough when it comes to regulation. That was all my statements, all observations. I do think that a /HROEBG� a lot of regions in the world are very similar to where Europe was in the early days or maybe not academic days but in the commercial world and building these interconnects and having a free and open peering clause it will help you develop and develop over services and that was my observations and conclusions. Question. ?
AUDIENCE SPEAKER: Kurtis, you repeated that peering is not a good thing for dominant incumbent Telcos.
SPEAKER: I said it didn't have to be.
AUDIENCE SPEAKER: Yes. I would like to, especially considering there were potentially people from different kinds of market than the western European market that you are focusing on there that might be present, a smaller share of a larger market can be much more better beneficial so it can actually in many cases in developing markets be still a good thing for the dominant Telco as well.
SPEAKER: Completely agree. Anyone else? Did anyone actually catch the presentation?
AUDIENCE SPEAKER: Remarks on one of your first slides, I can confirm that there is still I think about 40 metres of that thick yellow cable in the basement that connects the CIIN, if you are interested I can offer you half a metre at the next RIPE meeting.
SPEAKER: I will greatly appreciate that offer. Thank you. All right. I think that was it. Thank you.
(Applause)
CHAIR: OK. And if there are no further questions we will go into the IXP updates, alphabetical order starting with A. AMS first followed by DE CIX, five minutes each, your time starts now and I will go and help. AMS IX.
Kim: Hello. My name is Kim. I am a member representative at AMS IX. In the next five minutes I will give you the regular update on AMS IX. OK. Thank you. As many of you saw, we had, we experience add large outage last night at Nick /KA*F, I won't touch on this during this presentation but you can go to the technical list for the update or if you have any questions, please contact Hank or others or from Nick /KA*F, she is also here. I don't know where she is, but� OK. In the back. Historic overview of AMS IX:
The orange line you can see that growth in number of members slowed down somewhat but after summer this picked up. You can also see that the amount of port, especially the 10 gig port, are more often exponential. We have seen diminished net growth in members. This is partly due to consolidations in the market, but the number of new members is also slowing down somewhat. However in, the last weeks we saw a tremendous increase in new members, currently we have 12 members� 12 new members in process. The 10 gig ports are still growing, keeps growing on a high pace and as we expected, the fast Internet ports and gig bit ports, there is no growth in them (gig a bit). Just some migration from the bottom up. (Migration).
This graph shows the connection speed distribution for AMS IX members. The vertical axis shows amount� the number� type of ports occurs. The green bars shows 1 gig connections or aggregates based on 1 gig and the purple bars shows the 10 Gig connections or aggregates based on 10 gig. Although the graph doesn't show it very well, we have one member with eight 10 gig port connection. This graph shows that 10 gig connection aggregates become more and more common and the number of aggregates become larger. And technical consequence for this might be that with the currently used axis switches emulates or R16 with 6410 gig ports less and less customers will connect on the same switch, result not guilty less and less local switching and a higher utilisation of the core switch.
We� we have added seven foundry Internet switches and five for cross connects since May '08. For those unfamiliar with our set up, we run two networks in parallel, one active and one back up, so in one moment in time either the red or the blue network is active. In case of a problem, an automatic fillover is trickled to the back up network without loss of sessions. Some people see certain� some people with� some people see link flaps or with certain layer 2 providers. If you have more questions about this, technical questions, you can contact Hank or Stephen or also here, Hank is sitting over here in the front and Stephen, also. This is the traffic volume statistics, we truly switch over more than half a terabyte per second, more commonly 533 gigabits per second during the daily peak. What we have seen over the summer is the peak remains the same; the volume was very high due to streaming of the owe limb picks, tour deFrance and soccer championship by the Dutch national television broadcast companies.
What is new? Equinix is now live. It took some more time than both parties wanted but the switch is now (E Q U I N IX) (lower case) it runs and the first connections are active. CAR IX, the Caribbean Internet Exchange to be expected live the end of 2008, and then we have the Internet site by PI service, this is under development. In case parties are interested in this service, please contact Hank with specifics of how would you like such a service you would like.
The calendar, AMS IX general meeting and technical meeting next month, the 26th of November, inclusive the party. May meeting is in 2009 time and place being announced.
The peering forums, GPF 4 in /PAUPBT can in a. This was start at 28th of January till the 30th. EPF 4, it's in September, time and place to be announced. We also present at RIPE, NANOG, APRICOT and the NOG. AMS IX party during RIPE 58� sorry, 58� 85, sorry. On Wednesday. Special events, the partner event, this is before the general and technical meeting on the 25th of November. Sponsored events, ISP go cart competition.
If you have any more questions, thank you for your attention in front. If you have any more questions, feel free to contact me or one of my colleagues or if you consider become a member and come and have a date with me. I am Kim, member relation, representative of AMS IX.
(Applause)
Wolfgang: Hello everyone, I am Wolfgang from the DE CIX management GmbH and as most of you know we are running three exchanges in Germany. First one is the DE CIX in Frankfurt, DE CIX international exchange. Some updates to the numbers. We since the last RIPE meeting, we added one co low, we added one switch and we see quite a number of port increase. The high number of 10 gig port increase is due to the topology change we made and to make this clear, I split it up on our update here the number customer ports and total ports. So we now have 157 customer ports in 10 gig and 297 ports in total. And we also cracked the 500 gig traffic area.
News: We have some new staff, Thomas and Bernhard joined us in engineering and Caroline joined us in event management. We still look for business development person, so if you know someone, send us an e mail.
In 2008 in total we are going to add about 60 new customers and more than half of them from eastern Europe, so there is still a huge potential of customers, we hope to get some even more from eastern Europe region. We are going to a public beta with our Sflow portal based on our network product, going to do this in November and each our customers will be able to log in and see some Sflow statistics on the portal. Actually, we are going to put this live on the next DE CIX technical meeting, which will be on the 13th of November at Fransfer castle.
In this presentation, I want to talk about the technical infrastructure because Arnold already has done that on Sunday, so if you have missed the presentation, it's on line, just have a look there or talk to Arnold if you want to know further details. One new site, Frankfurt 5, interaction site, mid of year with that, our total sites in Frankfurt are DE CIX are up to eleven.
Traffic graph. We see a huge increase after the holiday period here this year, and talking to other exchanges, they also saw that. We can't really explain what it has caused, if for example some web hoster went live with new data centre and pushing out content. We actually don't know. So if someone from the content hoster can just give us an indication if they have put up something new, just let us know.
Yes. One of the other exchanges, regional internet exchange is the WORK IX in Hamburg. Now 30 customers, plan to go up to 40. 100 percent up time, we have updated the website. Might have a look, have a new website up for this and going to connect some of the content providers like Google. Our newest baby is the ALP IX Munich we announced at the last RIPE meeting. It's now up and running. It's force 10 powered, it's located in Munich and if you are interested to join, the first six months of service will be free. Any questions? Thank you (applause)
CHAIR: Next presenter will be JPNAP. Did you upload your presentation or not? So technical support will change that. After that it will be me and then mam ex and then NetNod and then the cheque ex EURO IX update and talk about EURO IX competition and then with go for coffee, hopefully.
One of these things, I feel like I have to talk, I don't like awkward silences, it's like being a TV presenter and still waiting for them to start this event somewhere else at the moment, you have got to fill the time.
SPEAKER: Good afternoon. I am from JPNAP. Thank you very much for opportunity to make a presentation. This is my first presentation at EIX. So I would like to briefly explain or the overview of the JPNAP. The first, our company is is established in 1997 and JPNAP service itself started 2001. This company was established by Japanese largest providers IIJ and also content providers. And briefly explain that Japanese, the broadband situation. Still, number of broadband user increasing and maybe you are notice that portion of their cable TV users are lower than comparing to other countries. And� we have 29 million of the broadband users. And this year's topic is in the number of users are more than DSL users and this happened in the summer of this year. And this graph is the Japanese government, gathering information from the provider, ISPs and exchange providers and estimated the current, the Japanese Internet traffic. 879 gig bits per Sec at the peak /P RBGS estimated the total number of Japanese Internet traffic and 303.3 is our exchange point traffic. The total of exchange point traffic. And the total traffic still increasing, 1.2 times per year, OK.
And JPNAP. And we are commercial internet exchange in Japan and at any time based layer two services and we have 3 exchange points: Two in Tokyo and one in Osaka. But exchange point are not� is not interconnected, and so these three are independent exchange point, and Tokyo 2 is very brand new. We open this site this April. And we provide the 10 gig and 1 gig first user port and also a link aggregation and maybe I will make a presentation about IPv6 in Japan, exchange point. JPNAP is currently already IPv6 available. So the provider can connect not only IPv4 but also IPv6.
And we are the number� we provide of attended ports among the IXPs in Japan, the number of customers using 10 gig ports is 26, approximately we have 60 or 70 customers so 40 percent of user are using the 10 big ports. The number of 10 gig ports itself is over 70 so many providers are using the multiple 10 gig ports. Some are aggregated with link aggregation but some others are load balanced by themselves. And the traffic trend of the JPNAP. Tokyo one is largest exchange point of the JPNAP and currently we have 171 gigabits at peak. And Tokyo 2 started much more smaller, we have just over 8 gigs, OK? And Osaka we have approximately 45 gig. And total we have 222 gigs and so not so big as in AMS IX and DE CIX and Linx but in Asian area maybe we are the most traffic. And JPNAP architecture, this is conceptual one but we are� we are preparing our redundant switches and connected them with each other, and we provide a very highly reliable system. We provide a completely redundant. And of our customers within 10 gig and 1 gig ports, main and back up ports, so we provide one according to the one order, yes, we provide main port and back up so we provide two ports. And these two ports are switched by the� by using the Internet intelligent optical switch. This one is just so I can� this is not same as AMS IX or DE CIX are using but this one we are using this optical switch for customers' ports, and this one is just like� if the main port lights go down, they automatically, the switch changes to the back up port in tens of milliseconds, approximately 20 or 30 so the customer's routers that is not since that are linked down or if there is configuration is currently, so the BGP, the session does not go down, so the providers can enjoy very stable connection over the JPNAP switches. We provide VLAN services not only public but also� we can private segment using VLAN technique and we also provide peer watcher services which is the traffic graph for each customers. Our traffic is analysed using the Sflow and the customer can see the trends of each peer and how much the trend of each peer. But of course, the customers cannot see the others peering traffic. And as you may know, the JPNAP is first associate member of the EURO IX since 2005 and we are contributing our experiences just like Sflow analysis and kind of things. So thank you very much for your attention. This is a brief explanation of JPNAP. Thank you very much.
CHAIR: Next up is mike. No questions?
Mike: OK. Great. So I am Mike Hughes and the CCO of Linx and give you a kick update in the intervening six months since I last spoke to you. I wanted to say thanks for the JPNAP presentations, some of those stats on broadband penetration was amazing, 70 odd percent picking up some 12 million people, absolutely fantastic. So quick look back at 2008, last six months, where we are in terms of the current stats and vitals, what is going on in terms of new stuff so. Looking back at the last six months we have fairlior I had Sumner terms of technical issues, these are all resolved and worked out now. It was very months of pretty hard work to get to the bottom of everything but now it all seems to be good again, thankfully. We have expanded our staffing, we have got new engineering staff, new person in dedicated to service provision so that is just working with Linx participate, new changes and those kind of things. New role in public affairs working with Malcolm, we have got Christof members relation officer, help Malcolm raise awareness of public affairs issues that might affect the industry generally and coordinate efforts in that area and also got a new person in the sales team as well to help us deal with the ever growing number of members that come to Linx. We have got continued growth in traffic and members and obviously the other thing we have been doing is deploying new hardware to keep up with that so we have a Foundry MLX in our biggest site and the number of other high capacity Foundry switches have gone in over the course of the summer. The other thing we did do after the issues that we had in the summer was with Foundry we did a specialist stress test scenario lab, built� what it will look like at the end of the year because we are just about to make a whole bunch of changes to bring up the new sites so it was important that we could prove that the architecture is completely sound and be able to do any tuning and also put upcoming versions of software that are going to be released soon or very recently released on to the platform and then we could actually check for bugs and make sure that these are fixed before we are in a position where we need to run a new version of software.
Quick picture, lots of cables and switches and gear in there. Don't touch any of those otherwise it will all break. And that was very, very successful week and hopefully we can go back and da that regular ace basis. Approximately 335 gigasecond peak, just like most other people we saw this funny flat bit during the year and it all picked up again in August which seems to be a very much common theme just shy of 300 connected members now. And quite a lot of (shy) capacity coming from 10 gig ports, 137 member facing ports, that doesn't include the backbone connections between the switches, it's over 200 if you include that. And that gives us something in the region of 1.6 terabits per second of connected capacity although not all that is being used. In terms of new features we have been running our Sflow portal for a while, we recently did an upgrade of that earlier this year. Extreme data is going to be available soon, we have done some upgrades on the extreme platform to be able to export Sflow data into the portal and we have new features for members, a lot of people said it's great to go and look at graphs, could I set that data out and put it into my own managing system what we have done is created an XML interface in beat at that at the moment and that is an authentic interface with a scheme that allows the back end database that drives the Sflow data portal to be queried that, allows members to take Mcto Mcpeer to peer type accounting data and proDes it themselves off line, that removes the need for us to do a lot of pre processing. Managed inter site private interconnect. We have been using this between most sites what we are going to do between the biggest sites is replace this which will either give you a 1 gig or 10 gig channel over a system that we are operating ourselves. Just be in the four largest sites for now and see how things grow and develop before we decide to deploy else where. We are going where the demand is clear. Opened three new POPs, none of the them in the immediate local, one of them in the city of London, we have got Equinix over in Slough, Telecity, live now inter and ex Quinn and Telecity will be available later this month once the file and build is complete. Both Foundry and� diverse fibre routing all the normal good stuff you are used to. Here is some pictures complete with pretty cabling, enjoy it while you can, you know what happens in reality. And where they are in terms of a map, well the Telecity and Telehouse sites are over on the far left, the yellow push pins and then green push pins are the new sites, so it's increasing our geographic diversity, getting us out of the docklands area and hopefully that is a good thing for a lot of people going forward.
In terms of the transmission capability we just did a single dark fibre loop runs through all the new locations. Both platforms sharing the same glass, we are pussing passive /PHUBGSs in, express port, we can put another 16 channels on if we wish to. ZRD, XFP optics, we are taking 40 kilometre or 80 /K*EUL and putting those straight into the switches and that basically obviously reduces cost and gives us a lot more flexible. There is no regem required which is interesting, I will step back a slide. Slough is way over here on the left hand side so there is a very long fibre circuit that comes all the way from the city all the way out to Slough, that is about 60 odd kilometres in terms of the distance that the fibre takes. It's about 20 odd kilometres as the crow flies but once you include all the routing of fibre and, it doesn't necessarily follow a straight line and it has to follow either road or railway lines and there is rights of way. That increases the distance quite a lot. 60 odd kilometres and what we did find we had to make sure all the fibre connections were clean so you learn a lot about connector loss and insertion loss of each connector type and patches and things like that. That alone is an interesting topic and maybe talk about it some other time.
Other bit of good news prices are frozen for 2009 so we reduced our prices midyear and we are going to be able to maintain those for 2009 and if you have got any questions, there is me here and also these other people, fine looking people here most of which are in this room that you can direct your questions to and I think most of us are here all week or at least until Thursday afternoon.
Any questions? And if there is no questions I will get the next speaker which is NaMeX, are you ready?
Working for the exchange in Rome.
SPEAKER: Hello, yes, thanks, my name is /PHA*URS, it's very hard to come after LINX, but anyway. Members is growing and during the last year in Rome, now we are 28 members and also traffic increased, now we have roughly at 10 big bit traffic peak.
Also we have� we increase our carrier in the carrier room. We add two international carrier, /KO*EPBLG enter and data communication. NaMeX decide to start to give available level one interconnection service starting from next year.
Just a news: We installed Verisign Anycast replica J with replica of dotcom and.net and we have part of their organisation of the first Italian peering forum meeting that was held in piece is a this May together with Milan� from Florence and Turin. We are going to have our fifth meeting in Rome next November. We started with our first 10 gigabit port, there are three members now connected to the 10 gigabits port and we started to give to our members immediately service, maybe you know BG play, it's available on RIPE site also and developed to the one of the four university of Rome, the and we start to have our multicast peering in production. That is all. Thank you very much. Any questions?
SPEAKER: No questions?
AUDIENCE SPEAKER: Alexander. You mention it takes as an Italian exchange, something there is either not right or you mean exchange. You mentioned the takes for peering forum, not an Italian IXP, correct? What is the takes, is that the exchange in Zurich Switzerland or who.
SPEAKER: The four Italian exchange points in Italy organise it, a meeting altogether.
AUDIENCE SPEAKER: Turin.
AUDIENCE SPEAKER: TIX is the Tuscany internet exchange.
SPEAKER: If there are no further questions, thank you very much and we have got NetNod next. After that, the check exchange. So we are back on schedule.
SPEAKER: Good afternoon, everyone. I work for NetNod internet exchange in Sweden. I think most, well some of you know me, I know NetNod so I will give a very quick introduction and freely free if you have any other questions. So I think we are probably the oldest exchange in Europe. What makes us a little bit different is we located in these underground bunkers where we always provide other critical infrastructure services so official Swedish time TP test, one of the route server operators in the world. We have six exchanges points in five different cities, I think most of you have seen these on the ground bunker pictures it, I just included them for /TKORB add a little bit of graphics in here.
What is new at Ned nod? A few things: The European peering forum, we are very excited about joining AMS IX, DE CIX and Linx as one of the co organisers of the European� we will be hosting the next peering forum in September next year so we will be announcing the location fairly soon, hopefully, and we are looking forward to welcoming you all there and I think it will be an exciting one.
I thought I would mention some of IPv6 efforts, as well. Apart from having enabled IPv6 in all our services we have had some talks about, because there is quite a lot of experience and knowledge of IPv6 in the organisation, we organised v6 workshop earlier this year which was very successful, Kurtis, he gives quite a few duetorials around the world in IPv6 among one here at the RIPE meeting. We have also talked about ways we can share this information better and put up some of the information on our website too, to kind of help push a little bit for IPv6. We also see a fair bit of IPv6 peering at the exchange and that is something we'd like to encourage.
Something I mentioned before, but at that stage it was just kind of in the planning stage, was the VDM solution, we are currently in the final stages of evaluating a WDM solution in Stockholm and something we will deploy at the two co los, Telecity and interaction this means we will be able to lower the costs of connecting to the NetNod exchanges there so that is very exciting.
And talking about prices, in the last two years we have actually really tried to or we have managed to lower the fees every year and as new peers come on board, we look at further reducing the fees and we currently looking at a further reduction of fees in 2009 so we are happy to� I think Kurtis is crunching the numbers as we see, so hopefully we will be able to� I won't actually go into the details of connecting to NetNod because some of you have seen these a lot of times but I thought I would include it in the slide set just because that might be useful.
So traffic, we reached 100 Gig in 2007, like all the other exchanges or I guess like everyone, we see every /SEURPL we see a dip in traffic. And I think that is correlation between the weather and the traffic as well, so I think you can read from this graph that the summer 2008 was better than the summer 2007. But also like Wolfgang mentioned, since then we have seen quite a rapid increase and we are currently at 160 gig where we are at the moment. And I thought I would include the traffic stats for Stockholm which I think it's about 80 percent of all the total traffic.
Like I said, continued traffic growth, just like the other exchangeses, as well, we see continued increased demand for 10 Gig ports, a lot of our 1 gig customers have now moved to 10 Gig ports. Another thing that we found quite interesting and we were very happy to see is the very high amount of traffic per peer at the NetNod exchange so we are not the biggest exchange point in Europe but the amount of traffic that every peer, every participant generates on the exchange point is very high compared to the European average so it's very positive.
That was it. Short and sweet from me. Like I said, I will be here for the rest of the week. If you have any questions, if you are interested in connect to go NetNod.
AUDIENCE SPEAKER: What do you mean by your WDM (WDM) solution?
Kurtis: Well, basically, as you know in Stockholm, we are� all the NetNod locators are famous bunkers and most providers are not located in the bunkers but our switches are so today we have to full fibre from the bunkers to issue the /SWAOEUTS and price you pay dependent on distance, two categories as the city monopoly and they have inner city and outsider city price category, the neutral co los except the Monday� are classified as outer city and what we do is instead of having a bunch of fibre put in a WDM system to replace this tonnes of fibres and reclassify them as inner city. That is basically what we are doing with this system. It's just, Monday.ly is a very good thing� it's a way to get around this pricing scheme. If the monopoly changes might do something else. We haven't picked a winner yet
SPEAKER: Because most people are at these two co los it's given us very little flexibility in working with the prices and this allows us to reduce the prices for them.
Kurtis: People might not realise, significant part of the cost you pay in Stockholm is 5 of the monopoly.
SPEAKER: Any other questions? Thank you from me then.
(Applause)
CHAIR: We do have is a speaker.
SPEAKER: I am the board member of Nick CZ internet exchange point of Czech Republic. They presentation is run by Joseph /HAO*UPL inbut his daughter was born about a a month ago so he decide to limit his travelling so that is why I am here so and I hope I get as good a job as he was able to do.
As I said we are located in Czech Republic in Prague. We have about 4 point presence. Of course, in the other exchanges we are growing, we have about 84 new members� 84 members in total. We also connect the top level domains, we have a special pricing for top level domains so we have five in the connected. Totally, it's 135 ports and what is good news that the number of 10 gig ports is growing. We have currently 28 ports. And about quarter or more of our members have also IPv6 connectivity and exchange IPv6 traffic in the exchange point. We have five switches in those, four POPs, we add a new one from the last RIPE meeting and we also added a new service for our members, we have route servers and about a quarter of the members use that route server so they have open peering policy and then exchange all the routes available in and this number is quite growing, this is just I think in two month. So this is the current topology as I said. We have we have four POPs and five switches in that. All the POPs connected by fibres with 10 Gig bit speed so there is multiple fibres among those points of presence. As you can see from the graph, the fast Internet ports are slowly becoming a history and the number of 10 gig ports will be very quickly higher than the number of fast Internet ports and, yes, it's still growing, there is probably nothing very important. The currently, the traffic in the exchange point is about close to 60 gigabits. We doubled the traffic from the last year so we are still keeping the same speed. Actually, the remarks from� that the traffic is correlated to the weather / RBLGS I made a small research about that and there is really a strong statistical correlation between the weather and the traffic in the exchange point, it's quite interesting and it works during the summer period so people can leave their offices, so the same happened in the Czech Republic, since September traffic is growing again.
We are currently working on a few things. We have problems with the port density at our switch in CISCO 6500, and we have also lack of new fibre optics in Prague, so we plan to use DVD M and we also plan to introduce a new topology so we would like to display the core switches and the customer switches and that means that they will probably try to buy a new hardware and we are looking at some options, Foundry, force, or maybe Nexus so I would be very happy if people could talk with me about that. That is the issues we are working on and that is all from me. Thank you very much. (Applause)
SPEAKER: If there is no questions, Peter is going to talk to us about what is going on at PAIX over in the US. Is that a USB key. You owe me a drink for coming to the podium with USB key.
SPEAKER: Hello everybody, I am Peter Koen, I am with switch and data and we operate the PAIX which is the switch fabric within many of our sites, and I am just here for a quick update in the marketing department. So, (PA IX) are within switch and data, switch and data is the company that owns all the co location facilities. We have about 33 sites, I believe, in the US. I should have put up a map but I was in a little short notice, I didn't have one, and within those are six main sites that have the PAIX switch itself. It says 1118th 60 sudden Monday and New Jersey, that is in the New York City area. Then in the San Francisco Bay area, two sites in 200 paul and� those are in the Northern California area, Dallas, Texas, I think everybody knows where that is. The other Vienna, which is in Virginia, that is a Washington D. C. Area /SAOEUFPLT Seattle in the north west in Washington State right near Canada and then Atlanta, Georgia which is on the east coast right above Florida. So traffic in growth, traffic and hardware this year, we have grown about 112 percent year to date, we are above 100 gigs of traffic now. We did a technology refresh putting in new switches in various sites that is ongoing site by site as time goes on. That is going to continue in 2009, and Sflow is run not guilty all the sites for customers and there is a portal like everyone has for their customers and finally we have built a peering tool to help customers utilise the switches more efficiently for themselves. So that is about it. Questions? Don't all come up at once.
CHAIR: OK if there is no questions. Thank you very much Peter. (Applause) and finally Wolfgang from Vienna and then Serge will talk about EURO IX and maybe Cara and then we can go for coffee.
SPEAKER: I am from the real Vienna, Austria. Maybe you expected Christian to stand here, he is on vacation trip, I am representing this time. It's my first EIX Working Group and Christian told me to be one to two minute introduction. This is the reason I only have one slide. So Christian told you last time that we are relaxing connection policies with our remote peering, we are allowing content providers and this is what came out. We have noticeable increase of participants, remote peerings and dual connects after the policy relaxation. So after we pushed it out to the public, we felt that really people seeing that and we got a lot of requests, a lot of small requests actually, a lot of small peers that are going remote, trying out internet exchange increase to gigabit or something at the moment, a lot of 100 megabit remote connections. We have content providers coming to us and saying you know that is good, we see them, we have a good increase in content providers. I didn't put any graphs on there because you know everybody has the same graph, summer hole and I got another theory: I think people actually realising in summer that there is life next to computers so they would go out and do stuff, probably. Some people might not realise that there is life besides computers. But there is. I promise. So we have traffic growths since May 2008 for about 8 percent. This is quite remarkable for us. We are really proud on that. We are peaking at 20 gig, close to 20 gig. And we are waiting for the 20 gig break to come the next days because we have really seen an increase in the last few days. We have growing number of IPv6 connections, there are a lot of� new peers actually come and say OK we go dual stack, so we are going to run IPv4 and IPv6 besides each other which is really good, you see the trend. And 21 percent of all connections are capable of IPv6, which is, I think, something. And this is it. Thank you.
CHAIR: Any questions?
(Applause.)
CHAIR: No questions. OK. I have got one comment and that is like it's really interesting to see with you changing your connection policies, how that is definitely had an effect on the number of participants and things, that is quite an important metric.
SPEAKER: It really does and we have lost peers because they wanted to do remote peering and stuff like that. We asked them if they wanted to come back, they are not there yet but I am sure they will come by time, so there is really need
CHAIR: You have people withdraw from the exchange because they didn't want the expense of having equipment, they want odd to remote peer over some sort of transmission with route to somewhere else.
SPEAKER: What we still do not allow is exchanging or, you know, selling upstream through the exchange, this is one thing we will never allow. I guess, or at least something we do not support so if they do they do it on their own because we don't have 24/7 at internet exchange
CHAIR: You are honest about your capabilities and you say to people if do you this and you try and basically take all your transit over the exchange as well and it all goes wrong don't come crying to us.
SPEAKER: Right, right that is what we do. And this is one thing why I don't show you our traffic graph because it doesn't look as decent as the others, this might be one reason.
CHAIR: It's still up to the right which is better Nan most of the stock markets right now, up to the right.
SPEAKER: Definitely.
CHAIR: Thanks, Wolfgang.
AUDIENCE SPEAKER: From DE CIX, did you see an increase in operational issues since you introduced your new policy that means allowing remote connections?
SPEAKER: Yes, there were operational issues but, fortunately, all our preventative measures really helped preventing all those issues so we have seen remote providers that would loop the connection back to us but we have port security configured on all our ports and port security really immediately shut the port down so we didn't run into any troubles but this is because we prepared pretty well, I would say, so nobody was affected at /SR*EUBGS except for the peer himself.
AUDIENCE SPEAKER: From a customer perspective is it as safe as it was before?
SPEAKER: Yes.
AUDIENCE SPEAKER: OK.
CHAIR: That rounds up the IXP updates and I think we have got time for Serge before we go for coffee. Five minutes?
CHAIR: How long do you need. You are such a rock star.
SPEAKER: I can put on a Swedish accent. Let Kurtis do it. My name is Serge from the European Internet Exchange Association. We have been around for a few years and I think most of you know who we are. We have got a few new members, Lyon from France is now back with us, they took a nine month vacation and then decided that they couldn't run the exchange without us so they joined EURO IX again, a French customer back again. /R*EUBGS, the text in Iceland are a member, I am still waiting on the bank transfer but they have joined us. I have to check that one up. And now we have got Africa affiliated with us in the form of /KAOEU /KWROER GPX, so we are now looking at 46 affiliated IXPs around the world. As well two new patrons, these guys are financially supporting us, extreme networks finally with us and Telecity Group as well, so we have got six patrons so we are getting bigger.
Traffic, everybody has been talking about traffic. I go around looking at all the stats. This is something I am not supposed to show but I am anyway because it's static. This is something behind the members' pages and what it is is automated grab of 37 IXPs from around Europe so this is basically realtime stats and one benefit of this, it allows me to see the actual peak time of traffic going on even though this is running over three time zones so I think I think got it set up in CET, in Europe it hits around half past eight at night and lows at around 5:30. These 37 IXPs are all members of EURO IX make up a little bit over 90 percent of the total IXP traffic in Europe.
Everyone talked about the flat summer, I have thrown a little red line across the graph here. Yes, it has been unusually slow growth since about the end of January, I only measured peak traffic web I do these sort of statistics, the last Wednesday of every month, but I have been consistently doing that over the years and it seems to produce pretty good and relatively accurate, as you can see through the summer that was really hardly any growth and then we came into our normal summer drop and the summer pick up was quite late. I am going to go into a bit more detail here. As you see, the end of April, which is generally when the traffic starts to die down, to the end of August, when the traffic really picks up, was only 3 percent this year, which is unusually low; in previous years as you can see, being around 15 percent or more. What I did notice is that the traffic then did pick up quite rapidly later, probably about two weeks later than normal. I am not sure exactly why that is. I don't think the summer was any warmer. I know in Holland a lot of people came back from vacation a little bit later than unusual. Maybe some other things were happening in other parts of Europe. I think southern Europe had a similar situation. But looking at the statistics now, it's really heading up again so we could be making up a bit on that lost traffic over the year. This is from the report that I publish once a year, what I have done is categorise the IXP as little bit, the total that I have underline in the middle is the average growth from the end of August 2007 to end of August 2008 in peak IXP traffic, so the biggest growers was Eastern Europe, Poland I have included, Russia, Czech Republic, Budapest, these guys really saw some big growth. I also categorised it in larger and smaller IXPs and the large guys seem, even though seemed to be pushing up above the average as well. The ones not coming so great out of the picture was the southern European countries, Italy, Spain. However, again, this could be a little bit due to the fact that they had that later growth spurt but I don't think it would go much higher than that 30 percent that is indicated there in that graph.
Globally, again, the European statistics that I have got here, I think are very accurate. However, the US and Asian ones are probably made up of about 10 IXPs in the US and probably about the same number in Asia so they are not as accurate as the European ones but I think it's� it is a pretty representative of the trend. On slide you can see the growth mapped through the year and you can see US and Europe are pretty much mirroring each other, however the Europeans, the traffic seems to have dropped a hell of a lot more, around Christmas. New year and summer period and this graph only goes up to the end of August, I think if I did newer one European one would be pushing ahead of the. And our Asian IXPs tend to be constantly increasing, although this year I did notice for the first time there was a bit of a drop around the Christmas/new year period, which I don't often see in the Asian IXPs. As far as participants go there is about 4400 participants at the 105 IXPs in Europe today, around 26100 have you� you� she was talking about the average NetNod's participants around two and a half gig, European wise it's about 445 /PH*EG per participant per IXP. I tried to somehow graphically display the ASNs that are multiple IXPs and it was difficult to do and this was the best way. Probably have the and look at the /PHREPBGS, have a look if he exact details. But from the outer circle you can see that there is, sorry from the inner circle there is 344 ASNs that peer at IXPs in Europe and then moving out, you will see that there is 3 that peer at 10 and so on and so on and in fact there is one ASN I think it's cogent 174 that peers at around 22 IXPs that I know of. So all in all there is around 661 peer more than one IXP in Europe.
News: This is about my last slide, don't worry. The 13th forum is coming up and I am overwhelmed by how many IXPs we are going to have there, and more than� 40 IXPs represented so it's again a record. That is going to be happening in Geneva. All these graphs that you saw are available in the EURO IX report for 2008. It's on our website. If you are interested in any of the stuff you saw here, you can get a bit more detail out of that report. Film competition, I am not going to do that and my little girl just turned 2 the other day. That is it. Any questions? Nothing. OK. Well thank you. If anyone is interested in joining� really proud because for the first time ever, every single IXP that presented was a member of EURO IX or affiliated in some sense so it was really cool. If there happens to be an IXP happening out there, my name is Serge and come to me in the corridors. Thank you.
(Applause)
CHAIR: OK. Thanks very much Serge. That brings us to the end of this installment of the EURO IX� sorry EIX. I am just name dropping. The EIX Working Group for today, anyway. We are now going into coffee break. EIX will conclude tomorrow, second session, so after the coffee break, 11�o'clock in this room tomorrow, Wednesday, and Cara already chairing that and Cara's item regarding the EURO IX film competition will take us to the first item on the agenda tomorrow. Randy: Tomorrow can we hope to hear from some middle eastern exchange points?
CHAIR: Nobody has volunteered anything from the middle eastern exchanges so far. There was the� Monday, that was the sum total of everything we found from the region. So if anybody has got any last minute contributions from middle eastern exchanges please come and see Cara or myself and we will see if we can get them on the agenda. Otherwise, half an hour coffee break now. After the break, NCC services will be the Working Group in that room and that will be led by Kurtis. Back here for 4�p.m. and otherwise see you tomorrow, hopefully, for the conclusion of the Working Group.