Connect with RIPE 58 : Facebook Twitter dopplr del.icio.us RSS linkedin iCal agenda
 6 May '09 ‑ Side Hall 2pm

The EIX working group session commenced as follows: As follows:

CHAIR: Hello, everyone. I would like to start here. So if everyone can look for a seat, if you want to sit, then we can start out.

You are here at the EIX working group, the European Internet Exchanges Working Group that focuses on anything related to internet exchanges and some peering‑related stuff, as well.

Today, I am alone, I work for the Amsterdam Internet Exchange, normally Fergus Mchey is here, however he had a bit of an accident on Tuesday and he is not mobile so he couldn't fly out here, so it's just me for today. I hope we can work with that.

So our agenda for today, first we have a describe and scribe and minutes writer. Thanks for that. Is there anyone who wants to add something to the agenda that was posted? No last minute updates? OK. Good. Then that is settled.

I do have some additions for the agenda, though. In the update session we will also have Equinix, we will have INEX, Map of the Americas who are not listed here yet.

Then the first item on the agenda is the co‑chair vacancy. As many of, you know, maybe not everyone knows because you weren't there, but Mike stepped down in Dubai last time and we are working for selecting another co‑chair to help us out, it often happens that one of me or for Gus might not be there, like he is not here today and it's very useful to have a third chair as well.

We have the system where we have a main chair and two co‑chairs and Fergus is the main and mike was a co‑chair so we are look fog fill that seat.

Now, we have nomination that is have been announced on the mailing list. If anyone is interested you can nominate yourself or someone else, if you will, and that should go to the EIX milling list which is EIX Working Group, so EIX‑WG@RIPE.net and they should come in before 5:00 today and that is when the nominations will be closed and we can draw up ballot sheets for tomorrow for the elections.

What I also wanted to let you know is that voting by mail will be allowed and this is coming down to EIX@gmail.com. This was drawn up for just this election ‑‑ it's EIX /TKOP right@gmail.com and we will use the single transferrable vote system tomorrow and votes should be in before 10 a.m. tomorrow for the e‑mail votes and we will do the actual election votes in the meeting.

Now, I have a slide on what are the Working Group chair duties. Right. Why did we, we do this? The Working Group chairs and the policy development process, when I stood for chair it, wasn't fully ‑‑ I wasn't fully aware of all the duties and I was kind of ‑‑ because I don't follow the policy development process in much detail in my regular activities. I was kind of surprised by the amount of work, additional work that has to be done as a chair. So, to make this very clear, we developed a couple of slides to give you some more detail and if you are interested in running you should know about this. And maybe it's interesting for the rest of the people, as well, so that you know what is involved, in fact.

So the policy development process at RIPE is a very well developed process, it's been there for years and it's been updated regularly. And the chairs have a very, well, a special duty there; in fact, what the chairs do mainly is to see whether the policy development process was followed, as it should, and whether consensus has been reached on the actual policy. And during the Working Group process.

So here it is. If you want to see this, it's in on the website and you can look it up in detail but basically, at the time of the ‑‑ after the whole policy has been developed and has reached consensus within the Working Group, it's shot through to the Working Group chairs and they have to review whether the process was followed correctly. Whether consensus has been reached and whether the could chairs are ‑‑ yes, I think that consensus has been reached, the policy is in fact finalised, and if not, it can go back and back through the process.

So that means as a Working Group chair you should be aware of the policies that are in process and follow all of that, which are in fact policies of the other Working Groups, as well.

So, you review the process and, yes, after last call, and then after about, in a number of days, mostly seven‑day window, consensus is or is not defined. And as a Working Group chair you can also still influence the actual content of the policy, but that is in the Working Group itself, so that is not in your own Working Group, unless it's a policy that comes out of your own Working Group, and then you should add that to your ‑‑ to the Working Group that actually develops a process.

Any questions about that? OK.

So we just wanted to /THAEUBG this very clear, so known is surprised by the amount of work that a chair has to take on (make).

So if anyone is interested, announce your intentions to run to the list and you will be added to the list. Currently, we have the ‑‑ the people that have announced their interest are Andy David son, Wolfgang Tremmel and Remco from Equinix so, those three are already running. Anyone added will be ‑‑ will be accepted and added to the ballot sheets that we will distribute tomorrow morning. The voting will be done by the single transferrable vote system and Ronan perry will help me out counting the votes there, he is very familiar with the system and has done it before.

OK. Then we go to the IXP updates and to start off with my own colleague, Steven Bakker from AMS‑IX.

STEVEN BAKKER: I am here to give a quick update on what AMS‑IX has been up to in the past eight months I think since last meeting. These are some growth figures. We gained members, we lost a couple of gig e‑ports, as mostly due to growth in the 10 gig area. And fast ethernet ports are just declining a little bit. Historical overview, not much really to add. The total number of ports is going do you know bit but that is mostly because people are hand not guilty multiple gig e‑ports for 10GE ports and ethernet are declining.

Architect and topology, we are in the middle of a big transition where we change our platform into all brocade /A*PL election stuff, and the base of the platform which is currently a double start topology with one side being active and one passive is replaced four course which is simultaneously and load balance over them and that will be done with VPLS and MPLS. For those who are going to be at our members' meeting on Friday, we will talk much more detail about these plans there.

Traffic and volume statistics: Traffic is still going up. We are just about, as you can see on the upper right there, we are getting into the summer dip now, which is an annual recurring event. But there is no real surprise there.

IPv6 is actually going very strong, even though it's about 0 .1 percent of our total traffic. There is ‑‑ has been a tenfold increase over the last year, so it's picking up. It's not doing a lot but it's definitely growing.

So, in summary what is new: We are going too far new access location, interaction AMS 5 which is in the /SKEUB /OL area. We are going to have that operational, planning is Q4 so before the end of the year we will be up and running there. As I said, the Newark tech tour and topology, come to the meeting Friday and we will talk more about that. There is some pricing developments, mostly prices going down, which will also be discussed on Friday. Small changes, we are still working on our new website and member portal and that is really in the last stages now.

We will be coming to the EPF 4 in September in Copenhagen so we will probably see you there.

Questions? It's actually readable. Any questions? These people, you can talk to and I will leave you with a last image of our typical kind of meetings. Any questions? No.

CHAIR: Wolfgang, you are up next.

WOLFGANG TREMMEL: Hello everybody, I am ‑‑ I was going to start without slides until he has finished setting it up. I am Wolfgang Tremmel of the DE‑CIX text in Frankfurt. I am running customer support (Internet Exchange) and I am going to show you the DE‑CIX updates as soon as we have the slides. Well if, it doesn't work just leave the icons on. I thought use ago PDF would make things easier than using PowerPoint, but obviously ‑‑ that is OK. That is the DE‑CIX update, I hope my five minutes start running now. So, some numbers, first. From last year, we increased our customer base quite a bit, we are now at 300 customers, but that includes the ‑‑ we serve eleven co‑locations and one thing you might notice is that we have changed the number of switches a bit; we, in 2008, the eight switches you see here includes the core switches and the edge switches, we now have seven and two core switches and the core switches are now Foundry MLX 32. We have 488 parts in total and of which are 21810 gig ports. (Ports). If you have a look at the port statistics, which I think it's the same as AMS‑IX 10 gig ports rising and 1 gig ports and fast Internet ports basically staying constant.

News: For the first time now on the customer ports, we have more 10 gig ports than gig e‑ports. Last year, we launched a better test of an Araber Sflow portal for customers so they can see statistics. This first better test has shown some strange results for some customers so we stopped that, /STHAOPD, went back to the drawing board and (stopped that) now I want to announce that we are ready to start a second, I am going to send out notification next week how customers can access the system and we are pretty confident that the statistics now show reality.

We replaced the core switches, we were using force 10 E 1200s, replaced with Foundry MLX 32, the reason for that we were looking for a switch which simply supports more 10 gig ports in one chassis and that was the obvious choice. Edge remains unchanged, customers are still connected on the force 10 switches and we are very happy with them.

News on the business development side:

We have a new employee on ‑‑ starting on June 1st, doing carrier relations. We are going to introduce a re‑seller programme in 2009 for co‑los and ISPs and our monthly necessary letter which I hope everybody of you is reading, has ask Dr. Peering section, which is DE‑CIX is the exclusive sponsor of IX sponsor of ask Dr. Peering.net. Check out the website. New customers: Well, Facebook inter fibre and Starnet, GCN. Facebook we broke our own record here, we had them up and running within 24 hours which was quite good.

Some recent events: We have been to St. Petersburg, Russia. I was there myself. We did a peering forum together with the Moscow Internet Exchange which I think was very successful. Basically, it was kind of half day like EPF, GPF, like, thing, and it was very successful, the slides are on‑line at the Moscow Internet Exchange, have a look and we mights be doing that again.

We have been to the balance cans capacity event and of course we are doing some round tables in cities where our customers can come to Frankfurt.

Upcoming events: There will be a DE‑CIX meeting in June, the date will be announced shortly. And on the 25th of August, there will be the big party, the summer party and before the party we are going to hold a peering workshop. From time to time, we get approached by customers who are really, really new to the peering world so we decided to put together a workshop, what is this peering actually and how can I use that? So, that is a half day event with talks introducing customers and potential customers from small ISPs, from the corporate world to the exciting world of peering and how they can benefit from it.

Traffic: Well, you can see the graph on our website, so not much new here.

Yes. That was it for DE‑CIX Frankfurt. But most of you know that we not only run an exchange in Frankfurt but also have two other regional exchanges, WORK‑IX in Hamburg, /STKWE quite successful customer round table there (we did). In January, 24 customers present. The traffic peaks at 2 .5 gig at the moment and we have 31 customers in total. Latest new one is /TKPWAOELG and Akamai is going to change us soon. (Google).

The other is ALP‑IX in Munich. We started that last year. We have now six customers up and running and as you see down there on the right side, the traffic is slowly increasing, we have about 300 Meg there at the moment which is not so bad for very new exchange. Join us the first six months are free, and you might see an announcement shortly ‑‑ we are awaiting more sites in Munich. Currently we are the Equinix near the railway station but we might extend that this year.

Any questions?

CHAIR: OK. Next up is Celia from Equinix. And next an announcement from Greece IX. Are you here? Right.

SPEAKER: I am from Equinix. I work in the group with John Taylor here and Remco over there and announced me a couple of hours ago I would be presenting so thanks for that. So, I wanted to give you update of the Internet Exchange in our facilities. Equinix things /THA*EUS having changes is great because it helps the customer to have services and here is the up dates. We have got eleven Internet Exchanges and in we have not ALP‑IX, you just heard an update about this. Is in our Munich data centre. AMS‑IX, same, you heard about it. C IXP ruin with Equinix in Geneva. DE‑CIX in our Frankfurt facilities. The Equinix does he will /TKOFRB, just installed and it's a successful exchange, few peering in place already. Linx is in our London 4 facility in Slough and you are going to hear an update later about Linx. ‑ and the just announce they can provide Linx from anywhere. The other in in Amsterdam is NL IX is quite successful with some traffic going on. So K LE Y R E X is exchange in Frankfurt 1, in our facility, they have actually about 100 members and they are present in over ten sites. /SWEUBGS IX in Zurich, they have ten locations including two at Equinix (Swiss) it's the best exchange in Zurich. Then we have got the Equinix exchange in Zurich in two locations, about 60 members, 6 gigs of traffic and 3 new members so far from the beginning of the year.

Equinix Paris, this exchange is now from two sites, our facility in Paris called Voltaire in Telehouse and there are already about 20 networks so the traffic is running up, so if you have any more question about it just come and see us later.

On a global scale, we are run a number of other exchange in US and Asia. Peak traffic is about 325 gigabits. And the other end we just updated the portal which is IX dot Equinix dotcom, you will find information about members present in exchanges in Europe, US and Asia pack and we also provide information on the network available from every site, so if you have a look, and tell fuss we need to make any updates on the network services reports, that would be good.

Finally questions, we are three, we are here until Friday, so feel free to ask us anything.

Thank you.

SPEAKER: Thank you


CHAIR: Anthonia here from Athens, Greek research and technology network, he has an announcement regarding new exchange in Athens. And next up will be Nick from ‑‑ Nick hill lard from INEX. (Hilliard).

SPEAKER: Hello, all. I work for the Greek research and technology network known as GRnet and I will present to you a new, let's say, project about the new Greek Internet Exchange which was created in 2009 in order facilitate the connection of the Greek ISPs. The new thing is that we are providing from now on a 10 gig bit ports to our peers, it's new, comparing the previous IX, Athens Internet Exchange and we are trustworthy entity which guarantees without discrimination open interconnection between Greek ISPses and from on totally independent from national telecommunications organisation and the the formal will eventually become obsolete.

Some of the members that are now joined the new ‑‑ this in effort are the Greek research fort, national telecommunications service for Greece. We are planning to move a K‑root instance, we run the former version for EIX now shortly to this new effort and hopefully all of Greek ISPs will become members of this new effort.

The new thing is as I said before, that we are now providing access to 10 gig bit Internet ports. This is very important for at least for Greece. And we also give to our customers traffic statistics, which is public ‑‑ for public access and Sflow based. Here there are two graphs showing the growth of GR‑IX and how AIX is becoming obsolete. We have a total traffic at around 16 ‑‑ 10 ‑‑ 16 bits, correct. One of our targets is we can of course some of the largest European IXes to peer with the Greek IX and of course drive down the cost of IP transit and more advanced and reliable services. We are providing IPv6 peering. We give a call to the domestic ISPs and we are waiting for their responses. That is it.

CHAIR: Thank you. Welcome here and I hope you will be joining us for the sessions in the coming years.

Next up was Nick from INEX and /THAFR Mike for Linx. (After that

Nick Hilliard: Good afternoon everybody. I am afraid I have no presentation this time. I thought I'd get away without doing anything at all up here but I was /PWHRAGD as somebody from the EIX Working Group usually does. We have the normal story of regular growth which seems to be entirely unaffected by the global recession, which is rather good. I am not going to bear you with figures, we have 180 percent growth over the past year, sorry 140 percent. In terms of interesting stuff, that is different from other Internet Exchanges, we now have about 55 percent of our exchange connect to go our root servers. There is some space at the end of the EIX Working Group to talk about root servers and we have some interesting observations. We haven't had the level of instability seen by other Internet operators, mostly that is due to a scaling issue but there may be one or two other reasons, as well.

We have several new interesting members, Google, we welcome to the exchange over the year /S* probably largest and best known and Google have just in the past two days started peering with our IPv6 root servers so we are expecting to see a little bit more IPv6 traffic on the exchanges as they spread their IPv6 love and goodness to everybody in Ireland.

Apart from that, it's business as usual. So thank you very much.


CHAIR: Thanks, Nick. So Mike is up now. And after that Andy from LONAP.

Andy: OK, I am mike Hughes, I am the CT O for Linx, I no longer have the yellow /TPHRU on my badge, as I resigned from being a Working Group chair, normally I would have been up here probably helping to run things but I have got other things to do.

So, I will till what has been going on in the last few months at Linx, particularly since Dubai. So that is the main sort of things that I am going to be covering today, Nick just alluded to route server stuff, that is quite an important thing at the moment (route server) no EIX presentation would be complete without graphs so here we have. We have added about 100 Gig of traffic in last six months and that seems to be continuing. The split between the two platforms, we have a Foundry platform and extreme platform and they are both operating and active in parallel. The split between that is is fairly constant, around 65/35 so the majority of traffic is on Foundry and the remainder on extreme. The graphs again are fairly similar to whatAMS‑IX is /SAOEFPLGT the 10 gig growth is circumstances 1 gig and 100 Meg are fairly flat, so when these bigger ports come out what we see is people buying more and more of the big ports and returning the smaller ones. Of course smaller members, new members come along that don't have as much track and they take up some 1 gig ports but basically means 1 gig is roughly flat N terms of connecttive capacity just over two terabits, it's the blue area on the graph here and that is the made up by 10 gig ports.

We were working on deploying new POPs, Equinix mention this had on Slough on the far left of the map so quite away from the existing sites, the existing are the yellow push pins on the right in docklands near London City Airport, we have got Telecity site in west London and the interaction site in the city in the financial district. That offers quite a lot of diversity from the currently ‑‑ previous docklands sites that most of the peering activity in London was concentrated in, so now there is opportunities outside of that. Going to Slough brought some of its own challengeses in the sense that gave us a very, very long metro span to work with, the fibre distance from Slough back to the POP in the city, bricklaying over 60 kilometres long, comes with its own challenges in terms of maintaining a reliable link, considering everything else was probably no more than 10 kilometres long or 20, so that put some new challenges in terms of fibre hygiene, that kind of thing, we are running DWDM on those Linx as well so we are doing with that with 80 kilometre optics and they are plugged straight into the switch.

To support the new site and to support the traffic growth we did some changes to our backbone topology on the Foundry platform, we remain with the ring topology, still using M R P for ring protection. We had to get the MLX type switches and make sure /TPHERP a ring with only the same. That allowed to us increase the make it greater than 80 gig, we have got 12 and 16 by 10 into switch Linx between the MLXes and then we have got RX rings connected off that. What that allows us to do is to right size the capacity into each site to be relevant for the actual demand that is in that site so we have got the MLX rings between the big sites and those are doing sort of anything between 8, 12, 16 by 10. And then we have got the RX rings which are doing anything between 2, 4 and 8 by 10.

Again that still gives us a good sort of stable topology with very rapid healing and reconvergence in the event of a network failure.

A quick map showing you what things look like now and the core ring is highlighted there on the right‑hand side.

And this is the extreme platform. Fairly similar, some slight changes just due architectural considerations in that platform being different.

What, we did was enlarged Telehouse POP, the original POP that was in Telehouse using on day‑to‑day basis was built in 96. The design criteria was for 100 to 150 members, we have now got over 300. Also the devices have got bigger, we are talking about MLX 32 switches and the size of ‑‑ one of the challenges was we could move somewhere else in Telehouse, there is a lot of existing building /KAEUL he willing and migrating over I think it was something like over 3 or 400 customer connections would have been a significant it is a income itself so it was fortunate that the a/SKRAEUS enter room became available and so we built /TPHAOUT that and took the wall down and that is what happened. If you imagine where that guy in the ladder is in the left‑hand pick /TAOURBGS there was a wall there and that was divided the two rooms, we have got double the amount of space there, that is allowing to us put in the bigger switches, in the bottom picture new cabinets with bigger switches in and that allows us to rebuild the old site so we are put not guilty 800 white cabs and they are much for dealing with density of cabling that is going into the switches. So route servers, Nick mention that had route serve issues are quite to the fore, really, in the IX arena, particularly larger. We have got two on each LAN both run Quaga, here is some usage statistics. And then the amount of people that are actually peering so some of the session that configure but there is no ‑‑ they are admin down and disabled. And then we have some v6 stats at the bottom as well. We have got quite a lot of peers particularly on the Foundry platform. We have hit some scaleability issues in Quaga because it's single threaded and BGP scanning is quite greedly, try and bring up as many sessions as it can and they seem to do things like cause race conditions. There are some bugs, if you remove a peer the the BGP might crash or trips max prefix and that is just related to some poor locking of events in the work queue in Quaga. Some temporary measures have gone into place, but all they are doing is buying us time. So what do we do and I think other exchangeses are look at this? You look at other implementations, are they going to get us out of the hole. Contribute changes to Quaga if we can and consider sponsoring further development and look at improving the IXP features and we were using in EURO‑IX meeting in Prague, I think we will go into this in a lot more detail. But for now, what we are doing at Linx is backing several horses in this race, we are not saying we need to fix Quaga or different route, we are evaluating several different options at the moment because you have got to and I think you have got to have a number of choices and see which one bares fruit.

IPv6 workshop was really well attended, we have got video of the presentations, they are on‑line, on our website at the URL at the bottom of this slide and they are publically available. It's really interesting, well really attended and real operational experience being shared. So if you guys are interested in hearing about how other people are deploying v6, go and grab the copy of the videos. Anybody got any questions, there is a lot of us here, the majority of Richard's team is here this week so if anybody has got any questions feel free, or grab one of us during the course of the week. Any questions? No. Thank you.

CHAIR: Thanks, Mike.


Andy: Thank you very much. This is an update LONAP Internet Exchange based in London. The company and the organisation is neutral, we don't compete with our members, it's not for profit, we are run for the benefit of the community. And our community run and help us govern the exchange as well. We have around 89 connected networks and another ten in the connection process so we are going to hit a magic milestone soon. The exchange itself is CISCO 6,500 based, 4 switches in a ring and a spare. Here is is a complete list of outages over the last couple of years. This is because we have some technical ‑‑ very strict technical policies that govern cleanliness and hygiene on the exchange and it's led to a successful policy, we feel.

The traffic that we do, we now hit 10 gig bits on most good days, and our graphs are up and to the right as they are meant to be. The sort of projects that we are working on operationly, we are also involved in the route server debate, because we haven't had any serious outages we want to have some so we are going to build some route servers so for the same problems that other people are. We think we are resolved, we probably feel Quaga is the wrong choice for us because there is lots of options including in London so in terms of our diversity model and message, we are looking at possibly running two route servers on different families altogether and having a lot of conversations with people this week to approve the case, the technical and business case for that. We have been loved over the last 18 months to look at expansion sites in the London region. It's actually quite interesting, it's not a case of doing a deal, buying a switch and configuring it, a /T* takes a long time to get the technical and the commercial situation right for any new site so we are evaluating ‑‑ hopefully by the next meeting there will be announcements and also we are find ‑‑ when we talk to people about connecting for the first time they are looking at who they have got to peer with and they are saying, I think we are going to need 10 gig port and we are looking at how to improve the density of our offering and grow traffic on the exchange that way. It's a brief update, I am here with some of my colleagues so please look out for us and if you don't wanton ask any questions in person, drop us an e‑mail. We will be glad to talk to you some more.


CHAIR: Mauritzio.

SPEAKER: Hello. My name is Mauritzio from NaMeX. The short update about our exchange point, our exchange point is based in roam, member based nonprofit consortium. We have 30 members, that is the same number that we had in the last RIPE meeting in Dubai (Rome).

Our traffic, we are around Gigabit peak traffic. In our data centre in Italy we have the major ISPs present in Italy and as you can see the traffic is more or less stable after the big increase that we have every year at the end of the summer.

We are reduced our fees in respect to 2008, about 20 percent, and we are trying to move from a traffic based fee that is sort of tradition in Italy to, a port based free, this is much more European. (Free). This include, this fee includes free additional port for link aggregation, and redundancy, two separate peering switches, double start topology and this includes free co‑location, so the members could install in our space up to ‑‑ up to 2 equipments, included at the same price. We started this year private VLANs for transit.

This year, we also started a carrier connection, meet me room service in our site. 15 carriers in this moment. These are one of the best connected sites in Italy. We are hosted in data centre owned and managed by university consortium, is university computer centre. They have a very high standard for the running the data centre, despite it they are a university environment, because they are very good maintenance and the data centre is really, really carrier great.

This site in fact is also chosen to host other critical infrastructure in Italy, especially from the national public network and government. (Chosen to) /*PL NaMeX is looking for a new site but this is not easy in, Rome has no data centre or at Vest very few neutral data centre. That is the problem. In Rome, we had just , so Casper is our centre, we use it in the past but this means that should be room for business for neutral co‑location company maybe. So if anyone is interested in Italy with this business?

In this moment we are working together with Casper to open a new /SAOEUFPLT we are trying to use data centre inside the university in Rome, we have four universities, most of the time they don't have the level of availability and quality that we look for.

We offer to our members additional service as, you know, the usual F root and IBG play that I know here in the RIPE community and we are closing an agreement with ISC, Internet Software Consortium to host our training courses for Europe. This is offered to our members, the possibility to have a very high level, quality course in Italy and especially in Rome.

Meetings: We organise two meetings every year. They have more than 100 attendties. This is the best place to meet CEO and CTO of our members. And the members like it very much because as usually, they are able to share ideas and to meet themselves. We have a general meeting, a peering workshop. The general meeting usually is in fall or winter and peering workshop is in spring or summer so the next one will be at the June 26th and will be our fifth peering workshop, a peering workshop is something more technical than the other one and if somebody is interested in participating in our meeting, just write to ‑‑ at this e‑mail address. (Participating in)

We are going to host the next EURO IX forum. So I finish. I look forward to see new Rome, and thank you very much.


CHAIR: Next up, Kurtis and after that it's ‑‑ Andrei are you doing the presentation? You are.

Kurtis: Thanks. I am Kurtis, I am the CEO of NetNod, normally we have my colleague here doing a much better job than me but she is at meeting in Brussels, she can't be here.

About NetNod, a small ‑‑ something we are doing just for, well, today and until lunchtime tomorrow, a number of the large Swedish operators the parent companies and us are being exercised in a disaster exercise and we are taking part with all the NetNod staff that is here as well so since this morning we have been fed these fake news events and there has been a number of critical scenarios in Sweden, we have been fed the false alarms, interaction with all the other Telcos and by this time all of gotten berg is without any telecommunications in this game. It's quite an interesting game, we have fake journalists calling me roughly every 30 minutes but that is one of the exercises we are doing just to check readiness and preparedness for the Telco networks in Sweden and this goes on until tomorrow.

Who are we? Nonprofit neutral exchange. Its processing with KTM network, started in 92. We are current operate exchange points in five cities, two switches in Stockholm and these exchange points, why we have so many is partly because of the spread out of the country, the length of the country, roughly 30 milliseconds north to south. NetNod has also established the idea that was this could act as a resilient infrastructure for the country, and these cities are supposed to act as back‑ups for each other originally. The actual switches located at government‑run bunkers and that hasn't been person connecting to us, you heard me talk about this before. We run a number of critical services, we are run of the route server operators, Anycast number of TLDs, 15 countries, some other things.

When you connect to NetNod we have this old scheme you heard me talk about this before, when you connect in Stockholm we will provide the fibres to your location where you are from the bunkers as they are not being disclosed and in the we get two connections, there is two switches and they are not interconnected, two times the ports, two times 1 gig ports and two times 10 gig. In other cities no fibre provider as there is in is in Stockholm, you will have to buy and coordinate it with us.

Some news since I saw you last time. As of January 1st we have lowered 1 gig prices with I think it was around 20 percent. New prices. And these are the port costs so, adding on the fibre prices, the Stockholm price is still higher but the other cities will be a bit lower. We will also be the host of the next European peering forum in Copenhagen, together with Linx, AMS‑IX and DE‑CIX, from the 14th to 15th of September. I will hope to see you all there. We are also going to run our NetNod meeting, we had one in February, we are running the second one this year, the same week as the EPF, I think that is a Monday, Tuesday and we are doing a NetNod event the nurse or Friday in Stockholm of that week.

Also, as you might know, we have a NetNod runs the exchange points in Sweden, a daughter company that has been used for a lot of DNS services there is historical reasons why there are two companies. Createsed quite a bit of confusion at times and we have decided to unify the name and as of sometime this year only start using the name NetNod to make it less confusing. We are going to go down the road of Lowe map, going to install a route server as well is and join in the fun. When we gave this update in Dubai we said to get around this inner/outer pricing, we are going to install VDM equipment at the two most popular neutral co‑los, Telecity and interaction and get a rate of this artificial higher price. We will have a unified pricing for those locations and the inner city.

Number of customers, have increased. We are seeing, like everyone else, clear shift to 10 gig ports. And we also see quite a few new providers joining us. Traffic volumes have been steadily going up towards 200 gig, there was a small drop a while ago. And we see that was for all the ‑‑ first graph is for ‑‑ the first graph is for all the changes, all the five cities and the second one is for Stockholm where we see this traffic drop as well. Any questions?

AUDIENCE SPEAKER: Why is the drop off (inaudible).

SPEAKER: , I kind of guessed that question coming. The ‑‑ so first of all, we don't run any S flows statistics, we don't know anything about the traffic going across the exchange point, we have no idea what type of traffic this is. And we don't like to speculate about reasons behind various events, we can comment when we know about them for facts but we don't like to make speculations. This traffic has happened on April 1st when it was also ‑‑ the date when Sweden implemented the EU directive on international property rights. The rights holder association and every newspaper in the world have decided that that was the reason why traffic dropped. We simply don't know. I can't say they are right, I can't say they are wrong. We don't know. I do know, though, that we in theory, I guess we seen increase in traffic days leading up to this we didn't. Also one of the most popular, ‑ located in Sweden, also said that they never saw a decrease in number of peers and the only I think that happened was, as you might have known on April 15th, was it? When the court case against the settled a number of other /TO*RPB trackers closed down but that was only two weeks later. There was a presentation yesterday by PhD student to had access to deep packet inspection data who claims the decrease was less peer to peer traffic that can be imagined but I haven't seen the presentation, it's not made public so I can't comment on this but there is a number of factors that seem to indicate because this wasn't all peer to peer traffic but I have no idea, I don't want to speculate, I simply don't know. But the world news press seems to know a lot more than I do.

CHAIR: They always do. OK. And after Joseph, it's ‑‑ /THAFR it's Josh for note at that and after that we have Maurice dean who wants to say something about Paris developments and we have ‑‑ VIX is the last of the up dates and then we start with the EURO‑IX update and announcement regarding IPv6 work.

Joseph: Going to give update, which is located in Prague. Since last RIPE meeting we saw increase of networks connected to exchange point by 10 networks, the total number of ports connected increased by 17. So this is our traffic. We saw peak at 86 gigabits and we are expecting 100 gig peak very soon. As for IPv6 traffic, there are more networks interested in that. We saw in late March peak around 117 megabits and currently 40 networks are doing this peering.

For our situation, we just, we are looking for new hardware, we want to decide in the second half of this year for new hardware for upgrade. Then we are working on implementation of the DWDM because there is a problem with new fibres between last week we hosted together with dot C Z the 14th EURO‑IX forum in Prague. So that is from my side. If you have any questions, I am available. Thank you. (DWDM)

AUDIENCE SPEAKER: Mark. The question is what is so high IPv6 traffic? Do you see ‑‑ did you see what traffic may be peer to peer traffic or web traffic or something else?

SPEAKER: I don't have any information about that. I can't say it now.

AUDIENCE SPEAKER: Czech Republic. I want to ask you what you are using for BGP? Already had issues with Quaga so it's interesting what you are using in your EIX.

SPEAKER: You mean for the route server?

AUDIENCE SPEAKER: Yes, of course.

SPEAKER: I am not sure, I think we have one, Cisco, and the second one is ‑‑ Andrei, can you help me?

AUDIENCE SPEAKER: Andrei: Currently we are running two Quagas and working on replacing one of them.


CHAIR: OK. Thanks for that.


Next up Josh.

CHAIR: Do you have any slides? You do.

Josh: Thank you. I am going to talk about Map of the Americas and some other exchanges. Wave data centres all over the world. So, Map of the Americas in Miami, is is a 75,000 square metre site. It's big building, handling 95 percent of the traffic between north and South America so if you need to get to Latin America it's the only gateway possible. We are at 170 plus gigs of traffic right now, I haven't checked in about three weeks because we don't pub /THRAEURB publically but it's growing. Almost all 10 gig connections turning up now. That is slide I accidentally put in. So right now we are looking at doing an axis 7,000 upgrade, going to kiss sew, we are the first one to use force 10 switches anywhere, have been very happy with them, about six years now, so it's time to look at something new. 51510 gig ports on a single chassis. There is a lot of fee tier problems they have. But they are going to write lots of bulky code to screw that up, we are testing it right now heavily and by 2010 we will put these in place, so they are really big if you can tell.

So little stats on different exchanges we have, say Paul low is about 15 gigs of traffic. Relate to to Miami over 9,000 PNIs. So we have force 10, 300 switches down there, much later, we will probably do that in Miami first. Bogota, you have to have that it's going from the airport it's pretty interesting. That data /SEPB tear has been successful transit 500 dollars Meg there. Quite expensive, so peering ports there are going to be relatively taken up pretty quickly and at least from a carriers in that low region. Dominican Republic, smaller, 800 dollars a Meg. Hopefully are /HR* stay on the island, have some good traffic there. In mad rid we are Hispanics, we are almost out of space, there is a few little racks left and that is it, that place is sold out. I think the next is interaction so once we are sold out please go to inter ah not until we are sold out, but then go over there.

That is it. Any questions? I did mine in 60 seconds. That was pretty good wasn't it.

CHAIR: If there are no questions, then in fact, Harald is next.


Harald: Hello everybody. My name is Harald Michl and I work for Vienna university. Is operating the Vienna Internet Exchange as a service and that is the reason why I am here now.

I don't want to talk much about statistics. The last year at Vienna Internet Exchange is described easily as business as usual, just a few values. We have now a peak rate of 5 minutes average daily of about 25 gigabit per second and since the last RIPE meeting we had an increase of 10 gig bits ports of about 50 percent that. Might sound a big value but it was from eight to 12 ports. So, we have nice little exchange point in the middle of Europe and, therefore, we don't have any problems with switch ports, with traffic exhaustion and something like that. It's just nice and fine.

A few words I want to spend now is about the projects we want to do in the near future which are with regard to the implementation of route servers. Vienna Internet Exchange point is not big with regard to traffic but we have 109 networks connected, most of them are very little regionalise ISPs and they don't have so much main power for configuring the network and therefore would appreciate a central organised route server. We want to implement one and the question is which kind of route server would be capable of doing that job. The features that has to be supported are IPv4 and IPv6 of course as well as four byte S numbers as we have heard that you had they are coming and (today) we want to give especially the bigger players on the exchange point to filter their announcement to different other peers who are not ‑‑ the small ones don't care about that because they take what the get, the bigger the ISP gets the bigger interest in fittering. We do not know which ports should be /TAEUPBG. We have already heard Quaga has some limitations and the two others we are considering are open BGBT and BIRD. The filtering modes that are common currently are taking the BGP routes with communities which works fine for 16 bit IS numbers but won't work for 32 bit IS numbers and our idea would be to take the information stored in the routing databases but we have had a look on what this documented in the RIPE database and what is real traffic on Vienna Internet Exchange and we had a very high rate of mismatches so maybe this isn't that good as well.

Very easy for us to implement filtering would be on our web interface that a participate pant can look in and say I want default peer yes or no. It's interesting for a new peer that is coming and I select exceptions from that so everyone accepts this and that. But of course that is a basis that works only on a peer to peer basis but not on single prefixes. So we would suggest then the peers to establish bilateral peerings for special cases.

CHAIR: Questions? OK. So that was the last of the IX updates. And last for today ‑‑ that is not exactly right, we have an announcement but the last presentation for today is the Euro‑IX update and after the announcement there is a coffee break.

SPEAKER: Serge from Euro‑IX. Just before I start, I just wanted to take a quick note because I organise a lot of meetings and I know how difficult it is especially socialising that the badges all flip over and Camilla came up with the idea of doing it on two sides so /KEFRPB see each other. So that was my little plug.

I wanted to do a quick survey of the room to see if everyone is paying attention and to see how many people have EURO‑IX. Can I get you to raise your hand in the air if you have heard of Euro‑IX before. Just keep it up for a minute. Can you raise your other hand in the air if it's the second time you have heard of Euro‑IX. Keep your hands up. OK. Keep them up for the rest of the presentation. It was to get the blood circulation going and /TOEFRB pay attention for a minute. (Everyone to) EURO‑IX we are an association of IXPs, that is how clever we were. Originally we started in Europe but now we are pretty much on most continents. The majority of our members are in Europe, 34 of them. But now we also have a member in Africa, Egypt, five in Asia, the one in India, our latest associate member to join. They have got I think seven separate IXPs in India and are planning on starting up another two. We also have three in Japan and one in Nepal and we have three in the US so really everywhere. Other affiliates are the trait reason tos, switch vendors and housing co‑low operators. Against the ones I have underlined are the two newest, Juniper Networks and interaction have joined us lately. They don't only financially support Euro‑IX but they join in discussion at the forums and on our mailing list. So this is our family.

One of the things we do is look at all IXPs, not just our member but all over Europe. So this is my best effort at, and I think probably the most accurate figures you will see as far as total IXP traffic goes in Europe. I calculated at the end of April it's about 2 and 3 quarter terabits per second which is around one terabyte up or 62 percent on the same time last year. So the couple of dips that you see around here is the summer vacations in Europe when all 400 million people take two weeks off at the same time and this is Christmas/New Year's Eve period /P‑RBG the stats are only taken at the end of the month. And this is due to the fact that I am the one who has to scan through all the websites. Looking back on 2008 as a year, I drew a red line across the screen just to show you pretty much the increase from the end of January through to the end of July, and you will see that there wasn't really an incri in the first six to seven months it, wasn't until around August that the 50 percent growth over 2008 occurred. Looking back at 2007, it's a bit down on the figures of 2007, 74 percent, and 2006 was also around the same figures as 2008, so that is a big of a look at the year.

Just to compare the start of 2009 to the start of 2008, I have drawn a line between end of January figures so these ‑‑ this is about the time when everyone is back from their winter skiing whatever they do in the mountains there, to the end of April increases. So in 2008, again it was a slow start, there was even a drop in traffic around end of February and there was an increase of only 3 percent over those 3 to 4 months. This year, we saw a bit of an extra increase in February and the increase looks to be around 6 so it's about on par or you could say it's doubled compared to 2008 but again a slow start to the year.

As far as our members go, the majority of them supply me with a URL which they update every five minutes, I scrape the them, put them in our database and we come pile a graph which pretty much gives us a realtime view of IXP traffic in Europe so this is obviously taking into account time zone differences, so while you will see here that it maxes out about 2.3 terabits, I think peak Euro‑IX traffic is around 247 so that difference pretty much comes from the different time zones, obviously London a lot of traffic is peaking a little bit later or earlier and Russia has a lot of traffic, however you want to look at T but this gives us a view. These little dips are not the IXPs went down or anything, it's just we weren't able to connect to the server for five minutes or so. We don't make this publically available as a realtime graph, I just update you guys with it so. Again the low comes just before 6:00 CEST, and there you can see it's around 600 gig per second.

This is a graph that I showed last time, it's ‑‑ while the date is is a little bit out of date the /TPREUGS still roughly about the same, highest growth rates occurring in east earn Europe, at least what I am terming Eastern Europe is IXPs such as the Moscow Internet Exchange, NIX Z in Prague, Budapest, the Bucharest exchanges, there is a couple there and also in Poland and they are showing exceptionally high rates, well over the 50 percent that is showing up across the rest of Europe and of course the large IXPs, by large I am talking about more than 100 Gig of traffic, about 5 or 6 of those a little bit above average.

So, who is causing this traffic? There is about 4600 culprits throughout Europe, 109 IXPs. 2005 only peer of them at one exchange point and the rest peer multiple, in fact there is more than 20 ASNs today and there is 2, possibly 3 now peering at 20 different exchanges points across Europe. You can access this information by going to the EURO‑IX website, euro/resources will take to you this database. Now, what this database will allow to you do is search /TPOPB all 4600 of those ASNs, nut ASN as I have done on the first slide here and you can see exactly at which IXP that ASN is present. Alternatively you can put in the organisation name and then see what ASNs they are using at the different IXPs, you see KPN shows up about four times, four different ASNs. A couple of changes that I made since the last RIPE meeting, one is the fact I have opened up the database a little bit more. Originally I only allowed people to search on Euro‑IX member IXPs, now you can search on any any in Europe, so try it out. You can put in whatever you want up here, Linx AMS‑IX and you will get a list of all the S N As present. Another change I made was the link from the AS number goes straight through to the Peering DB if they have an entry in the Peering DB.

A little bit of news: We had our forum in Prague last week where about 35 IXPs turned up. These are by annual, they last over two days and there are different IXPs or different cities in Europe every time. Basically, we discuss a range of issues, IXP‑specific issues, whether it's technical or commercial, regulatory, whatever the IXPs want to discuss, we have sessions on. Our next forum as NaMeX pointed out will be in Rome. Now, that is member‑only meeting; however, if you are an IXP or if you represent an IXP that is interested in coming to one of these meetings in the future, contact me. A lot of information you see in this presentation came from the reports that I produced on annual basis. That is also available at Euro‑IX/resources. The Internet revealed is is a film competition that wasn't a Euro‑IX competition, it was a competition that a few IXPs wanted to start up and we helped them to get together. I think we will get an update on that tomorrow, maybe.

CHAIR: It depends.

SPEAKER: You will hear about it in the future. A little bit more news. I want go into too much detail here because we are still finishing off the details but basically Euro‑IX is looking to build on our IXP family even further both in Europe and other regions. We are looking to assist smaller IXPs in a few ways and also those IXPs that can't easily come to our meetings on an annual basis, we are also look to go start autopsy new membership category for them. So again if you are an IXP interested in joining Euro‑IX and you haven't done it in the past maybe some of these categories will interest you and I will go into the details face‑to‑face with you. The Euro‑IX BCP, I think Wolfgang Tremmel who is leading, this is almost nearly completion. Wolfgang will gave short presentation on that tomorrow so I will let him do that. And there has been so much talk about route servers that we decided at Euro‑IX to put together a database for our members so they can list what route servers they have got operating, deem ons, problems with it and so on. There is a lot of talk and discussion going on about that at Euro‑IX. I think that is about it. Again, if you are interested in joining Euro‑IX, come and see me, I will be around, I will be here at the rest of the meeting, I will be usually hanging ARIN the bar at most of the social events. (Around). Any questions?

CHAIR: Questions for Serge? No. OK. We have one announcement regarding a workshop later today.


The announcement for the workshop.

SPEAKER: My /STPHAEUPL Paul, we are having IPv6 peering session starting at 4:00 just after the coffee break at the end of the hall near the cloakroom. If you do have production IPv6 network and are interested to meet your peers or to meet new peers, please join us. Thanks.

CHAIR: OK. Thanks for that, good initiative. Then, I just wanted to ask the candidates have already made their candidacy for the election of co‑chair tomorrow, stand up. So that is Wolfgang, Andy and Remco. So those three gentlemen have already announced that they are running for the co‑chair. And everyone can, in the session tomorrow, can elect their favourite candidate. We also have remote voting available through mil after 5:00 when the announcements or the candidates nominees are finalised, I will send /AOUT mil to the list on thousand vote /HREBGically and when your vote has to be in and we will add those to the ballots when we are counting the votes. If you plan to be here tomorrow, don't vote by e‑mail; this is mainly for remote voters that can't be there but are watching the webcast and are interested in voting. So, watch your e‑mail. OK. Thanks for everyone. And see you tomorrow hopefully.

WOLFGANG TREMMEL: Question regarding the electronic voting. How do you prevent that someone votes by e‑mail and also hands /TPH‑PB a sheet of paper?

CHAIR: What we are going to do there is that we are obviously watching who sends in the votes and they have to come from like real e‑mail addresses and identifiable persons and if I see people in the room and obviously, yes, this is not a watertight, waterproof system but if I see people in the room that have voted and see they vote twice, then they are expelled, at least one of their votes.