Connect with RIPE 59 : Facebook Twitter dopplr RSS linkedin

These are unedited transcripts and may contain errors.

The EIX Working Group session commenced on the 8th of October, 2009, at 11 a.m. as follows:

CARA MASCINI: Good morning everyone. You are here in the ?? you are here in the EIX Working Group session, 11:00. We are going to get started right away because we have quite a lot of presentations. I welcome everyone. We are going to have 18 updates and presentations by Wolfgang. If we don't have enough time left, we will go over to the next session for that one.

I want to thank the scribe and the Jabber and the text support for this session, Fergal, Alex and Tim. And I would like to review the minutes right now of last time. Is there anyone that has any feedback on the minutes that are on?line? We had quite a long session last time with elections and route server panel and there was a lot said so if the record is correct, then that is great. But if there is feedback, I would like to know. Okay. Thanks for that. Then the minutes are approved.

And well go to the IXP reports, which is ?? I am starting. Everybody has five minutes for their report, we will be quick and go through it. In principle, in alphabetical order, with the non?European IXs at the end of it. So here we go. I will do this in three minutes, I guess. For those who don't me I am carry Cara Mascini and I work for AMS?IX, I do the marketing there and so that is why I am the person doing presentations most of the time.

AMS?IX update: A lot has happened in the last couple of months. There is a number of people here, myself, Stephen and Martin Pels, they are all there, technical guys here so Martin will be talking about a couple of our technical challenges over the last, over the summer and so if you have any questions about that, just keep them for a little bit and he will be on stage later on.

So, what happened: Since the last RIPE meeting, we had a lot of new members, new ports, we added over 40 ports on the platform with about 25 new members between then and now. So that is sizable, for us it's a regular growth and as you can see, fast E and Es are in the decline since we don't even sell them any more.

So architecture and topology from AMS?IX: That is familiar picture at least to the existing members in the room. For those who don't ?? weren't familiar with it, it's a red and blue star network, one of them was active at one time. If something happens in the network, we would failover the complete network. And especially the latter continue with quite a number of 10 gig members, it did trigger more and more instability at the customer side. So we wanted to change that.

Also, we needed to grow, and the switches we had were just about the biggest that were available, so we needed to come up with a plan and /H*EPBG did that.

So now we have this, which looks quite similar, as you can see there are now four switches in the middle and there is quite a lot of intelligence on top of that or below it, if you will. If you want to know really more details about this, this is the MPLS/VPLS platform introduced over summer. Maarten spoke about it in the plenary on summer, since quite a lot of people flew in on Monday you might have missed it. He is talking about the migration from old platform to new platform later on so lots more detail on this later on.

Then, the necessary statistics: 766 gigs on the five?minute average peak. Traffic is still growing exponentially as you can see here. This goes back to '98 and all the way to here. And obviously 24?hour and annual statistics. Something that we haven't showed a lot yet but it's becoming more and more interesting, this is the IPv6 statistics. They are peaking at about 1.8 gigs at the moment and since these are done in the sFlow environment, we lost a bit of the history, so there you go, this is the history and then you move to the other one here. You can see that the growth in 2008 was phenomenal, it was about 1,000 percent. But even for 2009 it's been 225 percent up to now.

What is new? So the platform is new, we will have new access site. We have a price decrease that you can find out about on our website. We have new connection agreements for old members. We are in the process of doing that. We have a couple of events, GPF 4 and half just after NANOG and our own general meeting in November, which will be at the new, new the new axis location in Schipol so easy to access if coming from abroad.

What is new on the technical side: The Quaga route servers are quite unstable currently and we are moving one of them to open BGPD, the new updated version of it, next week and hopefully we will introduce more stability there. If you want to know more details about that, talk to the guys. And we might actually have a bit of an update in the session by Nick later. And then there is a new website and member portal coming on?line in the next couple of weeks. Any questions? Okay.

Next one up is the Denmark Internet Exchange. Where are you?

ANDERS MUNDT DUE: Hello, this is a punishment you get for socialising, suddenly you are told come and give a talk on your first time here. My name is Anders Mundt Due. I am one of the guys operating the Danish Internet Exchange and, well, I think I will just dive into it right away. Since this is the first time here, I will not tell you traffic reports. I will try to tell you as little as possible about what we did the last year, although I will mention it. But I will tell you who we are and how we sort of got where we are and why we haven't gone any further.

It started off by, well, one of my problems is that I couldn't really get hold of all the people who know all the history because that makes it necessary for me to walk the hallways in our building and get some of the old guys out, and I can't get the old guys out while I am at parties and other stuff at RIPE, but we did start in the early '90s. It was more or less a coincidence that, at least I think it was more or less a coincidence that we have had several of the Danish ISP located where our main building is, so we had lots of cables in, lots of people in, and some people thought, well, we could put a switch up here, so we did, phenomenal speed, 10 megabit switch, it was astounding. Of course back then people had two megabit lines so there was no issues and it worked quite well. At first we just did it and then, later on, we found out that, well, perhaps it's a good thing that we are the ones doing it because in the '90s we had fights and struggles between the various commercial ISPs in Denmark, we had TD C, which some of you might know which is one of the larger companies, and as a provider, no one really liked them because they owned the entire market and had a monopoly in the '90s that monopoly got more or less destroyed and we looked for someone able to carry the load of holding an Internet Exchange and since no one really wanted to try to trust the TDC back then, well we said, well, we aren't really involved, we have got no economic thing going here, we just have the building, we pay the rent, and we might as well just put the switch up, and we did. So we see us as the neutral ground for commercial ISPs.

Having that neutral ground attracted more and more customers over the years and by now we have got, well, these are more or less a guess at what we have, I just pulled data from the web pages and pulled information from our switches, so we have about 45 connected networks spanning 56 active ports. A few of them will have redundant links, a few of them are 10 gigabit links and most of them are gigabit links and we have grown to that over the last, well, 15 years, slowly, as people saw the need for upgrading.

We had a problem in 2003. We had a major power outage in Denmark which turned, well it did cut the power to roughly a third of Denmark. It wasn't very popular with people because that meant that all the power to the Danish Internet Exchange went off, which means lots of Denmark was isolated. We ended up in, as I have written here, in pretty much every kind of news media. We had press people calling us, we had interviews with newspapers, we had what was all over the technical papers all over Denmark. Everybody said, why aren't you doing it, why didn't you have battery backups and diesel generators, why is the Danish Internet Exchange run so ridiculously primitively? And I talked to the management about it and say what should I reply to people and journalists when they call and tell them the truth that, people didn't want to pay for it. We asked for people would you pay for us to buy diesel generator for Internet Exchange. Why would you pay for to us get redundant power in the case of a power outage and they said no, they wouldn't pay for anything so we didn't install anything and in 2003 it really hit them.

The thing is that, fortunately, we are government business, so lots of government equipment also had a power outage so I think two years later, the government had given us permission and had given us money to actually install the battery backup and diesel backup so we are looking more like a professional Internet Exchange now in many people's eyes.

So far, there is the natural progression that you get from networks, start over 10 megabit and go over 100 and gigabits and now we have a few 10 gigabit connections. We are hearing talks here and there from Danish ISPs and on non?Danish ISPs that, well, we should do more. They wanted us to do more and perhaps advertise more to people, but we are not really do it for the advertisements, we are not doing it to get customers here; we are doing it as a volunteer service for the people who might be here. The fact that we are upgrading the hardware is enough people said there was enough need to upgrade. The fact that connections are coming to our site has switched from being router equipment hosted in our hosting facilities and instead of short or long hall fibres going directly to the Internet Exchange. That is just a way things progress, that is what happens when people suddenly, it suddenly gets cheaper to get a dark fibre to our location than it is to install our equipment and rent some hosting facilities.

And the final bit is, I am not sure what we are going to do in the future as an Internet Exchange. I hear people constantly, well, here and there people tell me that they want to do a competitive Internet Exchange in Denmark. They want to take over the role of being the Internet Exchange. We have large competitor right to the west as in NetNod, which is a large Internet Exchange, and of course, we can hear them breathing. That is the way it is. Everybody talks to me, says are you connected to the AMS?IX and I have to say no we are not. We have got no reason to be there. They say how about putting up multiple sites in Denmark where they could cheaply get to one of our other sites in Denmark? If one person requests it ten he tells me and I tell my boss and my boss asks me if I am stupid, why would we do that? If we have enough people requesting something from us then we can actually say, well, okay, we will use some of the governmental infrastructure to actually pull some of the traffic so that still is on neutral ground and is something lots of people request, but small requests from, well, no matter how big a customer it might be, it's still just one person requesting it to me and there is nothing I can do.

Okay. Two minutes left. Anybody got any kind of questions to me? Can I say something without without having any graphs to prove what I am saying, but I know currently our traffic is somewhere around 706 gigabit; I have always thought that was really, really small. It's like why is it not more and then I have talked to some of the people that are actually peering at our site and it's because everybody of the big customers have one gigabit lines because it's just a service that they do, it's to make it easy for people to get here and get a little bit of traffic and that has backfired on some people. We had this as, I wouldn't say free service; they have to pay for it to of course get connected, we have to have some funding for some of the equipment. And some people actually to the ?? took advantage of that and abused it and said let's get connected to the Internet Exchange, we just need five metres of cable, able to run a gigabit and then we can put up a hosting company, so we have one or two hosting companies actually growing up two or three metres from the Internet Exchange. And what it really means is that some will see it as we have been funding startup companies and some will see it as a bad thing that they have that ability. I think I see it as a good ability because they have forced the rest of the market to actually drop prices in Denmark.


WOLFGANG TREMMEL: Hello everybody. I am Wolfgang Tremmel from DE?CIX and this is the DE?CIX update for the EIX Working Group.

On the platform not much changed since the last meeting. We still have the two Foundries in the core and we have 10s at the edge, we have 12 with the latest addition interaction five. Where we do not have switch yet but we are just backholding that to the existing switches.

We changed our pricing at the beginning of 1st of October to give customers, existing customers, an incentive still ordering ports, we have for the first quarter of 2009 best price guarantee, so they get the best price either from the new or from the old prices. You might notice that we dropped the policy of giving a huge discount on the second, third, and so on port, so every port is now the same. We also dropped selling fast ethernet ports, instead for entry level product we now have shaped gigabit Internet port with 200 megabit with quite attractive price for smaller customers. The complete pricing scheme is available on our website.

Other news: We now have 18 employees on board. We have couple of new guys, one of them is here. Can you please stand up. Andreas is doing sales, if you want to upgrade or order a new port now just talk to him. We also started DE?CIX consulting, so if you plan to open a new exchange in any part of the world, talk to us. We are going to help you, but it's not for free.

Events: Well, you know most of them of course, we also at NANOG and GPF and we are co?hosting the CNX in Frankfurt in October, if interested, go to the website and register.

Time to show new graph like Cara did, our IPv6 graph also done with sFlow and I think it might be a good suggestion for future EIX meetings if everybody who could produce an IPv6 graph, to do so, and either instead showing it instead of the IPv4 graph or showing it additionly to see that actually something is happening out there.

Yes, we also operate two other exchanges, one of them is work?IX in Hamburg. We have a solid growth there. It's a slow growth but it's regional exchange so we have at the moment slightly above 3 gig. Recently one of the bigger content providers signed up so that gave us a bit of a push and we also doing local meetings there, so we are doing round tables for the customers of these regional exchanges, we had 30 attendees for the recent round table. If you are interested in appearing in Hamburg, just talk to us.

This is basically the only real new item we have here in the whole presentation. We also operate an exchange in Munich called ALP?IX and we are opening a new site and 155, for those of you peering in Munich this address might seem familiar. We have also have roundtables there, next one will be in November, so if you are interested, just let us know.

Thank you. Any questions?

CARA MASCINI: Thanks Wolfgang.

RAPHAEL MAUNIER: So working for in Paris, I am from France. We talk about project in Paris. So, if you remember last RIPE just update about Pheon?IX, we were a group of people working in the same project to fix peering in Paris. So we were Maurice Dean from Google, Christian Kaufmann, Nicolas Strina, Raphael Maunier and myself.

So, we started to talk about the existing peering point in Paris, so it is about 12 in Paris and none of them is more than 100 gig of track and most of the traffic is going to Amsterdam, Frankfurt and London. So we made a survey about one year ago for about 70 answers from people, and we decided that we are on the good way and we start to work with people from the outside, just to be the new exchange in Paris. It's not a new exchange, a new platform that will be...people for it.

So this is a recent survey from people in, most of the people was from Paris and most of them, the other one was from London and Amsterdam. So are you prepared to pay a market rate for your port? So more than 80% was agreed. So do you need support, because there is no support today in Paris except for My Friend from Sphinx. Obviously people need support.

So, where we are today: Today we are not individual any more so we are working with companies, so we are no part of ?? not funding members so we are funding new company so this will be from Google, Jaguar and neotelecom and just because we need someone, we can establish the first, let's say, infrastructure. So we change as Working Group we change from Pheon?IX so this is new, you can see on the PowerPoint and we decided with some lawyer to build some dedicated ?? we saw we can't do something as AMS?IX in France we decided to move to GIE, it's almost the same as AMS?IX but with AMS?IX and the company in one company so this is for Sisco issue as always.

What is the status today: We have defined some POPs. We already have the support from so we are in discussion with interaction, and Equinix one time. We already have the hardware. 1,200. We already have the fibres from Neo Telecom so we provide fibres to exchange point so we will be able to have fibres between all the carriers and with our interaction with Neo Telecom we shared the marketing budget and hire the people. So what are the next steps? We are going to deploy the first POPs in the coming weeks. This is in discussion with all the exchange ?? to deploy right now so we are waiting for the people to sign so it will be something like by the end of next week. Official launch will be in December because we have the French NOG in December, the second week in December and we are going to have some events, so we talk from both sides to get connected to the exchange. We are going to exchange to new POPs so we are talking about Equinix and maybe Illiate the new POP. And we are talking with some people from other exchange point and with FranceIX and next step, we are going to be alie to be EURO?IX member. So this is phase 1. We are going to deploy two core sites, we are going to connect all the other sites 1, 2, 5, tell red 1 and 2 site ? equipment. This is how we are going to build the exchange so this will be fibre ring but point to point using ? stuff we are going to be about another fibre.

So this is where we are on the Phase Two is finish the phase 1.

That is it.

CARA MASCINI: Any questions for Raphael?


CARLOS FRIACAS: Hi, I work for FCCN and one of our activities is Gigapics. You see we nearly don't have any change in the membership but traffic is somewhat growing. So there is also practically no changes in the members doing IPv6 over the exchange also.

So I have seen the Danish Internet Exchange presentation, so it's good to know we are not alone, so we have some similar problems to our nature. So like them, we are also seeing less space used so people are not installing routers next the exchange, they are mostly using backhold connections so we are also nevertheless improving the platform so we have added 10 gigabit ports and we are currently accepting new members so we are currently talking about three prospective new members. So 10 gigabit, so 36 ports over the exchange, we supply short reach and long reach optics from the 36, 7 are currently used and there is a lot of space for new members.

So this is the location, we are near the airport and like any other exchange, we have 24/7 support, so if you have any questions?

AUDIENCE SPEAKER: If there are no membership of port fees, how is the funded, I didn't understand that?

CARLOS FRIACAS: It's a bit different from the Danish because we are into the university; we are an enter LAN so we have some funding from the ministry to help in this activity.


CARA MASCINI: Nick Hilliard is up next.


NICK HILLIARD: Good morning everybody. My name is Nick Hilliard, head of operations at INEX. Just a couple of boring statistics. 42 full peering members, we now have all of our members active, previously we had just a couple of members who for one reason or another known only to themselves they would come on as full members and not bother taking peering ports. I don't quite understand why but it happened. That is fully resolved now which we are very happy with. 13 associate members who get to come along and partake in free beer which is good for them and us as well. Our traffic is growing, still very small, as I have noted previously, probably the reasons for this are that because we all speak English in Ireland, most of the peer?to?peer content that we have would be exchanged with British and American peers so we don't tend to get a lot of native language peer?to?peer traffic like you would get in other European countries. But the statistics are very much in the right direction, about 125 percent growth in the last year, which is quite good.

We have a lot of our membership base using the route searchers, 67 percent. We have been pushing it quite hard. We do recognise that not everybody is going to want to route, use the route servers but they are very good for a particular segment of our user base and our policy is use the route servers unless you have a good reason not to and there are good reasons not to, but if you don't know what they are, then they are precisely applicable, they are extremely good.

We have a new point of presence opening up in interaction Dublin. It was scheduled to be opened last week but unfortunately, due to equipment delays and that sort of thing we were going to be bringing it on?line next week instead. Previously, we had been using Cisco 6,500 switches which are quite good, really nice 1 gig platform but they have certain problems at 10 gigs. But we did a bit of soul?searching about cost and performance and features and all of the usual stuff and actually settled on fixed configuration boxes and the new interaction node is going to be using brocade turbo iron T I 24 X boxes and fast iron this gives us a good combination of the right sort of density and the right feature set at a pretty good price so it's a pretty good win all around.

WDM, we have a fibre ring between all of our three sites. Now of course because the turbo iron boxes are SFP plus they don't have enough power to drive 10 gig over long distances, so we have to use XFPs for that that means WDM and transmit kit for that and they are really nice and like their stuff. A lot of benefits, lower cost, power consumption, there is one technical requirement that we don't have quite yet although the hard care is capable and that is just we need the IPv6 layer 3 filtering. It's not a huge problem because the amount of IPv6 traffic that we have at the moment means that it's not going to be a serious problem right now. But it is going to come on?line. And we are fully aware that this design does not scale to all IXPs, we don't pretend it does. There is probably a bunch thinking what is what in the hell are they doing. We think this is a good solution for us. So these are the boxes, turbo iron up here, small and the fast iron edge X tool box.

Here is the build, I am sure probably most of you haven't actually seen the inside of a data centre before, but in case you are wondering what it looks like, this is a cabinet and this is a bag here and we have some tools as well. We do actually have electricity in Ireland this, comes as a matter of surprise to some people.

CARA MASCINI: 10 seconds left.

NICK HILLIARD: Here is is a map of Dublin. We are down here and down here at the moment and the new interaction node is up here. There is a couple of other data centres spotted around the place. As you can see, all kind of on the outskirts here. Unfortunately, it's a gastronomic desert. If you want a decent lunch you are out of luck. This is our fibre ring, it's 29 kilometre ring and we have take answer spur off into interaction. There is another path actually kind of going up in this direction in the other side of this industrial park up here. We don't have fibre on that yet and we are trying to beat up our fibre provider to get connectivity in via there because at the moment this is kind of going in and out on the same duct which obviously has, is of some concern to us.

And here is our network layout. So, this is ?? these are our two existing points of presence here. Here is interaction Dub 1 and Dub 2 which is going to be coming on?line in January and the way we've worked it out with brocade and interaction and everybody else is that we have got two sites kitted out for approximately 45 percent of the cost that we might have spent on kitting out a single site, so major win all around and we are very happy. There we go. Questions?

CARA MASCINI: Any questions for Nick?

AUDIENCE SPEAKER: Hi Nick. You've got a hybrid set up there with different models, what resilience protocol are you going to be using?

NICK HILLIARD: We are going to have to start out with spanning tree. We don't have an option on this at the moment. At some stage in the future, we would like ?? obviously we are going to be retiring the Cisco kit at some stage, we don't know quite when. But if we were ?? if they are ?? if those boxes are retired, we would probably like to move to /PROBG aid MRP, there is a minor problems in the turbo iron don't support yet, it is in the product code map and should be coming into the code base relatively shortly, so spanning tree right now, not ideal but there you go.

MIKE HUGHES: Mike Hughes, LINX. Just more of a comment than a question, and I actually like the idea you have gone away and doing something that is the right size for the POP, you have said it doesn't scale but you don't need it to so the reason I think a lot of, some exchanges when they have been set up they have failed because they have looked at the sort of 95 percent of our exchanges that need huge infrastructures and the diagram that Cara showed earlier and they go away and try and build that and organise in a place where there is only 30 participants or something so I think this is really good stuff. NICK HILLIARD: Okay. Thank you very much.


MIKE HUGHES: Hi everybody, I am CTO of LINX. And I am here with a bunch of the other staff from LINX and I will point them out. Quite update on what has been going on with us, we are one of these 59 percentile exchanges with crazy actech tour. Vital staffs, over half a terabit at peak, over two terabits of connected edge ports, 340 members, so you know really good healthy member count. And from over 30 different countries, maybe even over 40 now. Switches to host 10 interconnected locations and spread all around London and not just in docklands where most IXs were commonly located. So we have got stuff in docklands, in central London and west London and here is is a quick map. There is the stuff concentrated on the right?hand side, as you are looking at it. The yellow dots and those are the docklands data centres out near London City Airport and then there is the three green push pins and three new location that is we built in 2008, over at Slough and Telecity over in acton and the interaction facility over in the city in Bricklane. They are all proving very successful and it's great that we have got the ability to provide some geographical diversity because obviously these locations are drawing power from separate bits of the power grid and other redundancy options. It's great we can provide you guys with more choice. That is ten different locations now.

In terms of network overview, we have not really made any major changes in the past 12 months. It's the same dual diverse LAN that serve a LINX member or is in the exchange community is used to. Switch connectors done at multiples of 10 GE right size to the group of sites that are on that ring and it's a ring type architecture, the brocade line, that is done with MRP, they are in protocol and in the extreme platform that is done with EAPS which is their ring protection protocol. Both achieved pretty much the same thing which is very quick failover and management of redundant links.

IPv6, well, I think no slide at the moment, no slide at the moment is complete without any discussion of IPv6. Yes you can, you have been able to do it for quite a long time now, it's done native stack over the same network and same VLAN and native dual stack, route reflectors of which more later, support on IPv6, right now our supporting services which are web mail and DNS and things like that they haven't supported v6 because we have not needed to do it and we have we have had other things to do, we are going to be doing that we have now got good internal support in our internal enterprise network to do this so in the coming months we will make our external website and DNS and mail and all good things available over v6. And the other thing was earlier this year we ran IPv6 workshop that solely to distribute Clue and I think it did a pretty good job.

How much IPv6 traffic? I don't have that number at the moment. I will try and refresh that.

Biggest project that we have had on in the last 12 months other than building out those three additional locations was to rebuild our POP in Telehouse north, it's 12 years old, it was the, not the only LINX POP but the second one that we'd built when we were growing about 30 members, so this was sort of generation 2, so it was moving away from the sort of stackable small switches and we built a POP and this was about in 1996. So this is now 12 years old and it's having to host more and more equipment. Moving somewhere else in that Telehouse north building was a real challenge because all the cabling from all the members throughout the building was all coming up to this room in D FM 6, and trying to move that stuff somewhere else was going to be a really serious nightmare of a job. Fortunately, the member in the room next door was moving out so we just asked Telehouse if we could rent that room, took the wall out of the middle and started to rewhether so we are rebuilding with new cab etc. And ODFs and replacing all the power infrastructure as well so that has been quite a big job because I am going to paraphrase Maarten, it's rebuilding the aircraft in flight here, so we were having to migrate people from devices in the old site to on the new side, at least we didn't have to run new cabling to the member terminations and you guys to go out and actually move which socket your equipment was plugged into. We managed to do all the changes from our end which was obviously made it much smoother migration. That is going to be finished within the next few months, it's been a long time coming but we have got light at the end of the tunnel now. We needed the space to hold things like the MLX 32, and we have got some nice neat cabling up on the top right so we have been able to take all the experience we have had from putting in these dense switching over the years and rebuild with the facilities to be able to handle the cabling and power and everything else properly.

Route servers: It's a common theme amongst ISPs at the moment. We run Quaga as route servers and if people didn't know what it was, as well as a route serves it's an extinct subspecies of Zebra. Here is a picture of one. It's in London, in a museum, and it is stuffed like most people that are running Quaga at the moment. So it's probably the best thing for it. The scaling problems are largely due to it's single threaded nature the fact that everything is run to completion, but we do need to do is get some better route server there. We are heavily loved in the EURO?IX route server and the goal is to produce at least one and hopefully more viable options and whether that is a rejuvinated refreshed Quaga that can handle the scaling requirements or whether that is alternative software, both of those are acceptable alternatives. I am going to go back to something that Nick was talking about in routing, RIR toolset, and that was a really good example of the same things that afflict Quaga which was software written by network guys, and therefore, it's not really been approached from a software engineering point of view in terms of keeping the code clean and optimised. I think that is certainly one of the things that we have heard from the initial research into issues that have been had with Quaga. Things could be optimised a lot more. Hopefully, we can get some stuff out of here, and we can replace these.

DWDM, I spoke about this in Copenhagen, it's a logical extension to the local PNI offering, one and 10 gig between selected sites, used trance mode stuff so those are proving quite popular with the ISP market. Here is the connectivity map, it works with the existing ODFs and sites and we can hook you up really quickly. Pricing and everything is on the website. Work done on sFlow internal tools so we can do some modelling, it helps us with doing upgrades, things like that so we can actually see between each switch where all the traffic is coming into the other switches in the network. And the other thing we have also done is invested in test equipment, buying all this gear we needed to have an optical spectrum analyser so we could find faults and deploy quickly, that is nice new toy that is going to be really, really useful in delivering quality service and also we are doing some upgrades on our IXIA platform, it's a traffic generator and analyser, we normally been using that to do performance test switches and stuff it led to, we have now increased that so we can do stress test routers so that is very, very useful when looking at route servers and Quaga alternatives. 2010, plan for over 50 new members and we are going to move our head office up in Peter borough because the lease has come to a end there. You have got to take on staff.

We are hiring, if you do know anybody who is interested, you must have the right to live and work in the UK. We are 15 in November, we have got our member meeting coming up in November and if anybody has got any member row bill I can't in their archives, please let us know and if anybody has got any questions, I am here as are some of our of our staff and sat in the back left?hand corner of the room. Any questions? No. Thank you very much.


CARA MASCINI: Next up is Will.

WILL HARGRAVE: I am Will Hargrave and here with my colleague Andy David son who is a WG chair. Just a quick presentation here.

Usual stuff, we are still neutral not for profit, four sites for now, Cisco 6500 based, we have some traffic. We have been working on quite a few things since RIPE 58, our route servers 2 on diverse platforms have now been deployed in a slow start ramp up. We have both open BDP and BIRD software deployed there. The other thing we have been working on is looking at next generation platform. We are turning up more 10 gig. Ports in our existing platforms, the same as many others have been telling us is beginning to run out of port density. And that is going on to the commercial side of things, we want to keep our 10 gig. Costs low. This is increase popular product and the new switches are currently cheaper. Also working and talking some data centre operators about growing out to new sites in London. There is some opportunities there in the local area to where we are already.

Who are we here to talk to? Other IXPs. But I am glad to be here to talk to more people, other IXPs we like to collaborate on the route server project which you will hear more from Nick about later and any networks here that peer in London and we would be able to help you, we are going to have a social in November in London, if you would like to come along have a chat with us and drop us an e?mail.

AUDIENCE SPEAKER: I notice you had an 800 gigabit spike this morning.

WILL HARGRAVE: It will be some monitoring error, I will go and fix it.

AUDIENCE SPEAKER: Is that a tacit acknowledgment that you fiddle your statistics?

WILL HARGRAVE: I think it's SMP and Cisco thing. Yes, what can I say.

Anyone else? Okay. Thank you.


SIMONE MORANDINI: Good morning everybody, I am going to give you a short update on MIXA, we have had 10 new members in 2009 and so an average of about one member per month and among the others we have connected media set, that is a national TV broadcaster in Italy and after RIPE in 2007 this is the second national broadcaster that connects to MIX. Thanks to this, thanks also to some upgrades made by some already existing members at MIX, we have had growth over almost 10 gigabit in the last year with a peak of almost 44 gigabit. The most part of this traffic actually came after the summer. As you can see from this graph. And if you consider that about one year ago around 7 gigabit that moved away from the exchange, this growth of 16 gigabits, that is pretty remarkable for us.

We also notice sort of awakening on the IPv6 side because we have assigned 17 new addresses in the last year and this brings up to 25 members that are IPv6 ready on the exchange. And while before we had only one peering LAN for v6 members, now we introduced those who are backup peering LAN just in the same ways that happens for IPv4. Unfortunately, I don't have graphs to show IPv6 statistics but it's something I am currently working on.

Again, on technical side: We have completed the expansion project of our data centre. We are on our own at the centre, and the last meeting was the installation and replacement of a second generator in the network, and now we have a full symmetrical and redundant scheme that you can see in this picture. There are two completely separate and independent power supply lines that feed the data centre. And we are strongly encouraging our members to take advantage of this scheme, not to have any kind of problem on the electrical side.

For fees for 2009, there has been a genre reduction in the prices. We are not charging our members on a per port basis but on a per traffic basis and so this is why you see prices that start from a certain amount. And we generally decrease the prices for ?? while the cost for 10 gigabit port is practically the same as 2008 we introduced 20 percent discount on the second port.

Finally, since April we have dedicated person on the marketing and communication. Some of you already knew her in Prague. We have, now, cross promotion activity with LINX and NetNod basically to ease the joining of new mutual members, and internally we are pushing on outreach by issuing newsletters and we published our first annual report and we are working on restyling of logo and website. That is it. If you have any questions?

CARA MASCINI: Any questions? Okay.


Kurtis: Thank you. My name is Kurtis, I guess most of you have heard enough of me last night. I am the CEO of NetNod. First of all you might have notice that had we not only do dinners, we also have changed the logo as of this time and we also run exchange points actually. So, we are the largest nonprofit neutral exchange in Sweden, founded in '96. Goes back to the we run 6 exchange points in five cities, two switches in Stockholm, we have 55ish providers connected, I think it's a bit more actually. When NetNod was established as most of you know run the switches from secure government bunkers and the reason for that is NetNod are classified as national critical infrastructure from the beginning and that establishment of NetNod the Swedish thought keeping connectivity inside the country was quite critical and need for high availability. This gave us the task as port of not for profit, you try to to help run critical infrastructure, we run the route server operators and DNS services for quite a number of TLDs. We run the official time distribution at exchange points from the U DS national sometime scale in Sweden. We are also the measurement points for something called TP test ways consumer tool are consumers can measure and validate broadband connections and reference points of the five cities of exchange points so you can measure the performance between your computer and see how close you actually get.

We also run a copy of the RIPE route servers and have for quite some time there is very few people actually use it. Because of these bunkers we don't disclose the bunker locations based on government requirements, one to each bunker from us. In the others there is no monopoly as there is in Stockholm, you have to order from your provider of choice.

I have quite a long time saying we wanted to change a bit of this. The pricing on the monopoly in Stockholm is the dependent on the location, where you are, and so we have this inner and outer city pricing and the, this is led to the fact between the two most popular and two only carrier neutral co?locations in Stockholm which is interaction we have substantial amount of fibres run not guilty parallel for higher price and we thought this was a bit silly so we have we plan for quite a while to do VDM, the I installation is done this summer we purchased the same trance mode system as mike was talking about. They are installed. We will bring the new customers we have pending life on this system and bring migration of existing customers. This migration will be done completely in the rooms of Telecity inter XION, you shouldn't have touch your equipment at all, it will be handled by us. Bring down pricing of 25 % move to the inner city pricing and those who are already existing customers will get this price from the date and onwards.

More update: We hosted EPN cope enHagan together with we are doing the route server installation hopefully this year. We have delayed this a bit, waiting for happy and joyful experiences of our colleagues and the work in other IXs, it can only get better as you saw from Mike's presentation. If there is some recommendations we are going to go for these and see what software is available.

We are also following up on Mike's presentation, we are also moving offices next week, there is going to be announcement going out, Wednesday/Thursday next week we will be busy with the office move and therefore we might have a bit of a delay in replying to e?mail, it doesn't affect operations, none of the operations are run out of the office space. Customers numbers: Five new members join and we have an additional five pending instalelation so this is the going to be the record growth year for us in terms of numbers of new members.

Traffic: Remember last RIPE meeting I got the question what did drop on April 1st was. We don't know. We don't keep records of this. We do know that some 60 plus journalist who is called us seemed to believe this was in relation to the direct that I have came into force on April 1, we don't know and don't care just shift bits. A lot of this traffic has grown back after the summer so after happened on April 1st wasn't something that seemed to have been sustained.

The same for Stockholm, which is clearly the largest city of all the traffic we have. We will see the same pattern.

And last, we also run I root and few TLDs around the world and we are number of exchange points and number of owe other locations. If you don't already peer with us, we really wanton peer with you so please send an e?mail and we will fix that too.

Questions? All right.


SPEAKER: Good morning, my name is Joseph. I am going to give short presentation about NIX.CZ which is located in Prague.

We have almost 100 members at this moment. We have 7 full?time employees, including the N OC working 24/7 and we are in top ten exchange points in Europe measured by traffic. We can look at the development of number of ports and networks and we see that 10 gig ports are really interesting for our members. As for the traffic, I will explain later that this year we experienced a little bit stable situations since spring, during summer. Now, I have the slide about the topology, so we have four locations now with five switches and we are preparing and planning upgrade because we are full of ports or we have almost useilised all four at next four location and we also implement ? between the port and we have today 4 times 10 gig interconnection between the POPs.

As for the traffic I said that this year we experienced growth since January, starting at 70 gigs in December/January, until April we saw growth but then the traffic somehow remained at the same level during the summer and now, finally, in September, the traffic started again growing and now we reached some peak this week around 97 gigabit in total.

So these are details graph. So the situation is not bad, the traffic is increasing. As for IPv6, I don't have graphs today, but I can say that we have almost one?third of members doing ?? operating IPv6 peering, 45 prefixes and all our services are available on native IPv6 addresses. About route servers: We implemented two route servers on different platforms, on different hardware, in different locations, one is BIRD and the second one is Quaga and, again, we can say that almost one?third of members is connected there so the peering is much more easier for newcomers and for new members. Here you can see the number of prefixes available through the route servers so it's almost 7,000 of prefixes at this moment. It would be better to have, let's say, some traffic statistics from this. Regarding the route servers, I have to say that my colleague will present tomorrow more information about BIRD, which is I would say good alternative to Quaga, and the presentation will be at the last session of RIPE plenary between 11:00 and 12:30 tomorrow.

About our future plans: We are expecting price decrease since January 2010. Then we are working on some hardware upgrade because we need to fulfil new requirements for members and also have to mention that the next RIPE meeting will be in Prague in May and NIX C Z together with C Z.N IC will act as host for that meeting so we are looking forward to seeing new Prague. So that is from my presentation. If you have any questions, I am here.


FRANCK SIMON: Good morning, I am working for the French ? and I am for this network. And one of the activity of is to run exchange point so we do have a national exchange point in Paris and we do few regional exchange points. So, regarding the SFINX, we have two POPs in Paris one in interaction and one in Telehouse 2, historic POPs and we do have now about 90 ISPs connected to the S F I NX and the traffic for each of these POP depends it may vary between 12 and 15 gigabits of traffic on each S F I NX POP and we do have, in fact, allocated ports, I should have written active ports. We have about 15 active ports for the 10 gigabit ports and more than 40 ports active for 100 Meg and 1 gig and for the 100 Meg ports indeed they are free ports.

For the we have two Cisco catalysts and we do have two dark fibres, we activate ?? two times 10 gig LAN dars on each fibre and they are fully redundant at optical layer, in fact the point today we have over the chassis almost used, I think we have only one slot free on chassis and so we plan to not buy some other cast list as users as you may have seen and exchange point we will buy some terra switches and we don't want to use any more catalyst switches because of the problems you already know.

In terms of services: During this summer we have migrated IPv6 and IPv4 in the same VLAN. More historic reason one for IPv4 and one for IPv6. We still have one dedicated VLAN for IP for multicast, IPv6 because router is national research network so we have been running IPv6 multicast in native mode for several years now so would be happy to establish IPv6 multicast peering it seems there is no huge request for multicast. And regarding DNS we have secondary service for .FR and we are going to migrate it to Anycast service to improve and we do host three instances of route servers, recently I root instance was deployed interaction 1 and for F and M route servers. Of course we support port aggregation and provide VLAN services where 24 hours N OC and as national network we do have computer emergency response team so we are able to deal with security issues on the SFINX itself. And we have just deploy a New Portal for I would say the management of the contact with ISPs to provide them access to their stats and so on.

For the upgrades: In fact it's more than upgrades. We are going to renew the first ? it's not that we are going to buy some new, we are going to remove the catalyst to buy terra switches for the two we have but an option to add a new POP and then I have written here either keep Kath list at edge for I would say 1 gig ports or less but in my mind I think I won't use them any more. I would prefer to buy some new switches rather than buying maintenance for all the switches that are any way too big even for 1 gig port and we better buy some new switches, maybe some smaller one only for 1 gig port and keep terra switches for 10 gig and more port N fact, we plan to buy this new equipment within the next few months, and so we will send the architecture that we plan in our mailing list in which all our ISPs are connected ?? sorry, we have a mailing list for all the ISPs connected so we will consult the ISPs to know which is the best solution for them because as I said, we are nonprofit organisation and we want to make sure that the money we collect from the ISPs is used in the best way possible so we need to consult the ISPs about the architecture. We are going to open by the end of this year new exchange point in Marseille; it will be an optical exchange point, if some people are interested to know more just drop me a mail or maybe we can have a discussion at the break.

AUDIENCE SPEAKER: Can you go back to the first slide, please. When you say there is no transit allowed, do you have anything in place that checks for that?

FRANCK SIMON: In fact, we do some stats, basic stats on the switches and so we just check that always entering on the switch is going out, so (and) in fact, there is no real I would say, no real technique or solution to check that. It's just that in the contract you do sign when you connect to it the SFINX you agree you don't do transit so each time we have seen stats that was a suspected transit, we directly contact the ISP and ask him to stop otherwise we shut down the port.

AUDIENCE SPEAKER: Thank you. Mass.

AUDIENCE SPEAKER: About your switches.
A. You said terra switches are you still in the selection process then.

FRANCK SIMON: Yes we are in in contact with other switch vendors but you ignore them and we will make our choice within the next few weeks.

CARA MASCINI: Thank you.


LUCA CICCHELLI: Good morning, ladies and gentlemen, I am TOPIX, first time for me here for people who don't know who we are and what we are doing, that is the first ?? probably one of you doesn't know anything about Turin, so where we are and remember last winter Olympic Games, yes, 3 years ago it was too much time for us. But regarding us, we are no profit consortium founded in 2002 and just as a first goal to create and manage and distribute IXP in the north and west of Italy. At the moment we are 12 people working there with more than 60 members and we are also integrated in our local government programme on broadband that is called W I?PI E, three years of our region. What we do: We do two things: Internet Exchange, of course, and everybody knows what we are doing; on the other side we are also doing the so?called development programme that is a programme to help in their start off phases providing them with Internet connectivity and hosting but of course I really focused on the Internet Exchange part. Our model is as I said before, at the Internet Exchanges. So starting from our infrastructure, we also, as I said before, the Internet Exchange provides bandwidth for the Internet ventures that are hosted in the development programme but the development programme on the other side helps us to build an open environment in which members and non?members can share their experiences and I told you before about integration, about the local government programme. Here is the network that is ?? the left ring, the right ring sorry, is what was provided by the local government programme, in such a way we are very distributed IXP. More details on the URL below.

How to access to topics: As we have in the region some very small ISPs we still have 10 MEGS ports but also providing 10 gig ports, at the moment only four but they are available only of course in Turin and how to connect there with your own infrastructure, of course leasing connectivity from carriers, you can see the most important nationals and also international carriers that are available on topics. But also we allow specialist ISPs to be connected in wireless, allowing them to put their antennas and ahow long them with cabling to the switches. Of course, the housing is available on each.

What we have done more is a connection with another small Internet Exchange in the closer French region in Lyon. Why? Because we are closing regions with common interests so we have decide today do something together. What we have done is a connection, layer 2 connection at the moment. Each IXP maintains its independence. What I mean? I mean that members in TOPIX remains only TOPIX members and they can peering together.

Just to be more quick. Our traffic was up to 15 gigabits per second but unfortunately IPv6 traffic on our exchange is more or less one percent. We have 40 autonomous system connected. You have seen before we have more than 60 members. The reason is that not every member is using TOPIX as an Internet Exchange but they are participating in the development programme, so most of them are doing both of two things, but not every member is using TOPIX as an Internet Exchange. Our platform is a mixed platform, especially in Cisco platform because we started with that but we have now platform also force 10. Of course only in the core. The edge is only Cisco. But as we are open we are trying this new technology and we are hoping in the future to try a new one, for example. We are multicast especially IPv6 enabled, we are exchanging some small IPv6 traffic and we are hosting F?root and J?root servers.

Just to give you a more detailed ?? more details about the core architecture. What is new about what I have mentioned before? Okay.

Finally, we have our route server, now with Quaga but my colleagues are looking for new one, the route server is now working on two separate hardware servers. My colleagues have developed a DB tool, NOC DB tool, as you have seen we have a complex, even if we have smaller complex architecture because we are doing different things so this happens, this tool helps us in network configuration in management and it's an open source tool and unfortunately tally only in Italian. We have deployed a connection with Padua. We talk about that.

The latest slide about the traffic. This is an example, a sample of the daily view and this is a sample of what is happening in a year. We have an up and down traffic, as you can see. Any questions?



We are discussing whether we want to move a couple of the updates to the afternoon session. Can I see a show of hands who is in favour? And who is against. That means we are moving a couple to the afternoon. So after VIX we will more to the afternoon.

MARTIN FISCHER: My name is Martin Fischer and I am working for the Vienna Internet Exchange and I will give you a short update on how things are going in Vienna and I will keep it really short because we are running quite out of time.

This is the kind of mandatory slide I have to show you. We are very close below the 25 gigabit of peering traffic. We are still locating on two sites, one at university and one at interaction Vienna with about 100 customers, the nearest Microsoft and Akamai, they brought us a lot of traffic so thank you acmy. The 100 customers are operating 140 ports, amongst them 45 IPv6, part of them v6 only and part of them just v6 enabled.

What we are working on, let's say what my colleague is working on, is establishing route servers. The work is now in progress and we hope to have beta test on?line about December this year. The route server will be based on BIRD. Wolfgang asked me twice to mention that the guys at NIX C Z were highly communicating with us and very fast with all the box fix so is thank you for that. There will be no BGP communities necessary because all the peering relations are built and written in our database, so the database is the source for peering traffic and it's managed through our customer portal where you can set all your peering behaviour and you can set all your peering data. There will be a site specific configuration to keep the cross traffic between our two sites very low so you can prefer routes and prefixes from your peering partner at your own site. There will be also the possibility to kind of pre penned ASs to configure the peering behaviour as you want to and, yes, hopefully there is this route server on?line in December and we are really looking forward to test it and hopefully it will work as we expect it to.

One last thing: There will be a technical workshop at the end of November in Vienna at the 25th of November, so if you are interested to come to Vienna and to participate at this workshop, just ask me for the details or just check our website.

If you have specific technical questions about BIRD and about how our route servers are built, feel free to ask my colleague directly. He will sure give you answers, more precise and helpful answers than I will be able to give you, but of course you can come to me and ask myself, I will just forward the questions to my colleagues. The very last thing, just a screen shot of how our route server will be manageable by our portal, so any further things in December this year, hopefully. Thank you.

CARA MASCINI: Any questions? No. Thanks.


That will be the last one for this session and we will do the rest of them in the afternoon when Andy is chairing.

RICCARDO LOSSELLI: I will try to be fast, I promise. I am one of the v6 exchange member so I will try to introduce you to the exchange without entering too much details as I am not one of the technical staff of the exchange.

The exchange started in June, while operationally just started in September so it's very young. You may think why another point in Italy as we already have other six? If there is any sense for it, we think yes. The region where we are, which is the eastern part of the country, quite the opposite of the TOPIX exchange has very high demand for broadband and has a very poor support from traditional broadband lines. This is the reason why in our area, we have about 25 major wireless ISPs and up to 50 minor ISPs and ? wants to promote the access to broadband to ISPs and also benefit them from peering on a local area instead of peering in Rome and Milan, even if the country is not so big it will say to local operators to peer some 300 to 500 kilometres away, of course increasing performances too. The idea is to keep the information local. Also, many of the operators are quite young and are not concerned too much yet to the quality of the service. Without giving any mandatory rule to follow, the idea is to develop some policy and service level agreement for the operators to follow to improve the quality of the service they are providing and to give services to the local government, too.

Why initial axes point in ?? one of the main research points which is the new ?? new generation network of the research ?? research network in Italy. We have a high number of operators, local operators in the area and there is a request for grid computing, and start ups that can benefit from the presence of the exchange and another interesting thing is that the local government was trying to find a solution to provide broadband and high quality service to local office it's and local administration without investing and spending too many money on communication infrastructure. The idea that show up was to give access to the market for every ISP in the area and while checking the quality of the service having the neutral access point to act as a hub for all the operators in the area. Another good thing is par with a is right in the middle of the diverse path from the central and south of Italy to the northern part of Italy and to the southeastern part of Europe, so Austria, Slovenia, hung Gary where we have a strong link on the economical and cultural side. That is pretty much the structure of the service. You can see we have about 10 members, four of the members are big Italian national operators and the others are pretty much local, the university and local government network connected to the exchange. We have exchange without neutral access point namely Turin as Luca said before. The exchanges managed basically by the local university with technical committee by the university and a technical, additional which is the operator committee formed by one member ?? one representative of each IXP member, basically the technical decisions are made between one representative of the technical committee of the university and all the members of the exchange. Of course there is the operation team which are the people that really work on the hardware.

Okay, the infrastructure and architecture is quite normal. Layer 2 infrastructure for peering and IPv4 address space dedicated to IPv4 peering. Unfortunately at the moment the exchange has no IPv6 support and this is, okay, this is something we really want to change. Of course the monitoring and so on. One last thing: The interconnection with TOPIX and this is the model of the over network which is the network from the local government. You can see that all the offices of local government can buy broadband Internet services from different ISPs so it does not interfere in any way in the market. It also opens the market to new operators and all the traffic is brought to the neutral access point in Padova which acts as a hub and allows the local government technical infrastructure to monitor the quality of the service that are delivered. Okay. That is it.


CARA MASCINI: Well, then that concludes our morning session. We hope to see you back in the afternoon. We don't start right after lunch but at 4:00 again, and Andy will be chairing that one, so if you are up for ?? we have three more updates then, we have session about Wolfgang, premier of the EURIX film competition, more with the reports and route server Working Group and then we will have Maarten Pels from AMS?IX talking about building ?? rebuilding a plane in mid?air and /KA*RT /T*BGS by Equinix and that will be the afternoon agenda. Thanks everyone. Enjoy your lunch.