Skip to main content


Note: Please be advised that this an edited version of the real-time captioning that was used during the RIPE 56 Meeting. In some cases it may be incomplete or inaccurate due to inaudible passages or transcription errors. It is posted as an aid to understanding the proceedings at the session, but it should not be treated as an authoritative record.

EIX Working Group, Session 1 11:00am, Thursday, 8 May 2008

CHAIR: Good morning everybody. Sorry, we are a bit late starting, address policy over ran a wee bit. We have got quite a busy day today. We have got about 14 presentations for the morning, and we have lots of swag, so we are going to have a quiz and [unclear] going to hand out the questionnaires and if you fill them in, we'll award swag at the beginning of the second half so you have a good reason to come back for that as well. We have [unclear] fleeces, T-shirts, hats, lighters, business card holders, espresso cups, pens, lots of things. Laser pointers.

So that's the morning session. Afternoon session, again, is orientated towards IXPs, I have a euro IX update. Update on 1,400 gigabit, some observations from apples IX on v6, recommend and Greg are going to be talking about the peering configuration survey they have been running for the past couple of months. Then we have got a couple of things which aren't on the agenda on line. One from Curtis from v6 working group session again on how they did their numbering at met nod and then a BCP proposal for quick discussion and mention it have from Greg on having a union BCP for v6.

We are going to be using minute cards. There will a five, a two, and a stop. So please [unclear] oh, and there is a one as well. So please if we are waving cards at you, please do stop.

Anyway, with we'll start off and our first person up is AMSIX.

CARA MASCINI: So, hi, I am with AMS-IX, and I'll give you a couple of the updates on what we are at at the moment. This is a historic overview, I won't go in there but I know people like to see the trend and all, so here it is. We are at about 500 ports, 400 gigas and 300 members. As you can see here, most of the growth comes from the 10G reports as you can see, that's pretty obvious for all of you. Since last RIPE meeting, 2007, September, that was the statistics I gave there, or at least I gave there. We have grown with 54 ports on 10 giga and the others are more or less basically stable.

This is the topology. As you can see we added a new site. We are in five sites now. EU networks in the middle. We started out to implement a core switch there but very soon was to obvious we needed /SPHOR space on the edge, so we quickly implemented an edge there as well. So this is what it looks like now. As always, the red and the blue star network, one of them active, the other one on standby and if anything happens in the active one, we'll swap over to the other one.

The traffic statistics. As I said 400 gigas, you can clearly see that it's night weather here, but not here but also in Amsterdam and around Europe, as there is a big drop in traffic, as always when the first sun comes up after winter, and of course, holiday season as well.

Well nothing much more to say about that. What's new? We implemented a member relations and support team who do non-technical support member relations but also contract and order management. They are very easy to find for our members and can be contacted through many, many channels including RCIM and all that. We know, since we were growing as an organisation, that we need today put emphasis on that, so there you go.

We have special high end access solution, a couple of the members are actually using them for aggregated 10G eports, and it's WDM based. And as said, one in extra locations EU networks which is already operational as you can see, and we'll add another site, we selected recently Equinix, the new Amsterdam. It's Amsterdam /TKO*UT east.

Further, we will extend the Sflow service that we have with raw RDD file downloads and native IPv6 to IPv6 graphs. And some other news is, some of you may have heard at the European the global peering forum, that we are going to help set up an Internet exchange in the the Caribbean. They ask us for some support and since it's Dutch oil and all, we started thinking about it and we are helping them set up the exchange. We are going to run it remotely from a technical point of view for a while and see how they do and hand it over as soon as we can.

If you have got any questions about that one, just please ask.

And the agenda, meetings, forums, events and all that.

Next event is going to be EPF for us in Dublin in September. And well, and our local barbecue and general meeting coming up soon.

That's it. No questions?


YVONNE HAHN: Hello, my name is Yvonne Hann, I am coming from the DECIX Internet Exchange and I am giving an update of our company.

This is my first English presentation, so please be be gentle.

First of all, some numbers, we increased or numbers of customers to 246 out of 33 countries. We now have ten colocations. I am going to talk about this later. We have now eight first-end switches, 381 gigaabit ports as well as 191, 10 gigabit ports. We are doing traffic and the peak traffic from around about 4/70 gigabit per second which is 100 gigabit per second more than last year.

News about our website: We have a new design and improved function alternatives. It was before, a regular web server, it is now a content management system. Customers can log in, coworkers can change passwords, they can do ordering by themselves, of course they can do this right now, but in future, they can do the invoicing by themselves, the target is that it becomes more self-administration abilities.

We have new staff. Stefan [unclear] could be, sitting over there, he is the junior network engineer and me myself, Yvonne, I am in the customer support and last but not least, we have our new DECIX girls on board since Tuesday.

What's new about our infrastructure? Our first interconnects are using DWDM transceivers, we have passive DWDM [unclear] and fibre protection devices. He replacement of the fibre protection devices is now ongoing. Additional first-end switches have been installed in DECIX 1 and DECIX 2 and the [unclear] peak flow flat form will provide detailed information about your traffic.

Yes, the greatest new in the last six months were that Equinix is now with us, DECIX enabled site number 10. It's a standalone building fenced in, in a fenced area and, yes, it's own property and they expect to have an increasing amount of customers within the next twelve months.

To give a view where Equinix is located, we have DECIX 1 and 2 in the east. DECIX 3 and 4 in the west, and DECIX BoF also in the east is Equinix located. In the middle you can see our DECIX office.

Other news. We are going to expend our office in Frankfurt, Frankfurt Competence Centre, therefore we need some additional staff. So if you are not happy with your current job... feel free to visit our website.

We have openings in business development and network engineering.

And I will now give the word to Frank, who is giving a short update about.

FRANK ORLOWSKI: I am not giving an update on the girls. I'll be talking about our to relatively new Internet exchanges, but is a Work IX in Hamburg, we took overwork IX about a year ago. So we have 30 customers up there. The platform is rock solid. We will start to have some more marketing campaigns around Work IX, so we are expecting around 40 customers before the end of this year. We do 10G E services in Hamburg. We have actually two extra sites in the city centre and well, we also have a new logo for Work IX which makes it look a bit like DECIX.

The next slide has some information about a new Internet exchange we are about to build in Munich, so about 25 minutes ago, we signed a deal with Equinix about deploying the Internet exchange in their Munich facility. It will be called the ALPIX, ready for service will be July 14th, 2008. It will be another force 10 powered EIX, we'll just do these services. The first six months of service will be free of charge. If you want to join, talk to us. Our idea is that from day one we'll start with ten customers, to get a bit of traffic on the platform and at the end of this year, we are expecting something between 20 and 30 customers.

Well, that's it from DECIX, Work IX and ALPIX.

ERIC TROYER: Hi, I am Eric Troyer, with Equinix, and I was educated in the United States and my English language skills are pretty bad too so work with me on the presentation.

Glad to be here and give this quick update on our company. Fergus said I had 30 seconds. I am sure I will exceed that by at least 600 percent. Many of you out here know Equinix probably because of one of two reasons: You are either appearing on our exchange fabrics in the US or you have heard of us through some recent acquisition activity in Europe. Namely, and firstly, IX Europe which is a name you are probably familiar with in the past. We acquired EIX Europe in September 2007 which gave us the red food print you are seeing here on the map. It's always hard to represent a map graphic with pinpoints on it without it looking evil like global domination take over but I haven't found a better way of doing it yet, so...

But in addition to that, in February of 2008, we also acquired a company called VIRTO and in addition to getting a really great data sent another Amsterdam, we also got Remco van Mook.

We love you Remco.

We have about 1,800 customers. Global company. We are in 18 different markets. Ten countries, 40 data centres and 280,000 plus square metres of space, so around 3 million square feet.

Real quick overview: Because Equinix is a US-based company, the US EIX landscape is different than Europe is that the data service providers generally operate the Internet exchanges and we are no exception to that rule. In the US rerun the largest Internet exchange in the DC metro, technically Ashbourne Virginia. We have 650 ports altogether and ten metros. Over 225 unique customers. That means unique entities. Someone lake Aid Limelight for example will be in multiple metros where we run Internet exchanges they will count as one unique customer. Generally what we see as far as traffic trend is our traffic grows 90 percent year over year in aggregate.

So a little focus on the EU strategy in particular. As a recovering peering coordinator who went to the dark side of the marketing and having been a customer of the incumbent ISPs in Europe our main charge in going into the EU was not to disrupt the very good work that the incumbent IXPs do, rather we want tonight part of their growth. So there was a specific decision made not to come in and launch our exchange IX service in certain key markets where there are incumbent IXPs, for example N Germany as other DECIX friends just announced or Frankfurt enabled the DECIX. In the Netherlands, by August 2008, we will be M6 enabled, and in the UK, we are in mature discussions right now with the London Internet exchange. However, in other markets, like France, for example, where we received customer feedback, we are proceeding with exchange. We have launched our exchange in Paris. Additionally, through the EIX Europe acquisition we are running what was formally the TIX, which is now the Equinix exchange which will be prebranded with the full exchange portal and Sflow and everything else involved. In Hong Kong, we are going to open an exchange before the end of the year. This will be the first commercial exchange alternative to what's in Hong Kong now. In Tokyo we are expanding into the shin [unclear] owe building to kind of give a larger footprint. Singapore we are seeing continued growth on the G C X platform as well as the exchange platform. In Sydney it's a mature market with very few new people coming on but mostly in the way of upgrades. New developments on the IX front in general. We are developing a new portal, which is going to have more features and functionality coming up before the end of the year and MLPA servers for the entire system. IPv6 we are seeing much more increase in traction. Across the board we are all dual stack in all US markets. If you are appearing on Equinix exchange and you don't have an IPv6 address, please contact us, we'd love to hook you up with that. We are going to be launching a GRX platform in DC similar to what the M6 runs in Amsterdam.

You can mail me at [email protected] if you have any questions.

NICK HILLIARD: Good afternoon, Nick Hilliard from INEX in Dublin. I am going to give awe brief update today. One slide. We are mature in our market at this stage in terms of membership growth it's relatively slowed down. And most of our growth is actually come through port upgrades and that sort of end of things and we have in fact seen a peak organic traffic growth of more than 50 percent since RIPE 55, which is pretty good.

We do a lot of video streaming, and this is proved that we, in Ireland, are obsessed with two things: Sport and politics and the politics is particularly good because for anybody who has been watching Irish politics, we have been having a regard eventful time with, well, corruption and resignation of prime ministers and all of this sort stuff so it's wonderful. We have a huge amount to be thankful for to our our Prime Minister recently. Unfortunately he is gone now, so we are looking for other areas to expand into.

Our technical operation; yours sincerely now, has now finally exceeded one fulltime employee, which is very good. I think it's a good indication of how things are going at the exchange.

We have done 10 gig upgrade recently in the past few months. Probably the only thing notable is that it was just pretty easy. It's plug and play. We have 10 gig customers in the exchange already and we are working on several more, which is good.

Just a quick note there. We are doing a unicast storm filters on 10 gigaports. We are currently doing multicast and broadcast storm filtering on every other port.

We have also recently implemented a route server cluster for those Americans in the audience, it's a route server. And the, this was an interesting project actually, and it's a little bit more subtle than it looks. I am hoping, I am trying to beat the EPF people into allowing me give a talk there about what we figured out and what was interesting there. So, if you are in Dublin, if you just happen to be in Dublin on you know sort of September 15th and 16th, you might hear a little bit more.

And, ITF is also going to be held in Dublin this year, so we are not officially involved with either of these two conferences, but we will certainly be having a presence there, and we would be very happy to talk to anyone who is in Dublin at around that time.

So. That's all. Any questions?

CHAIR: Thanks. Andy, it's you.

ANDY DAVIDSON: Morning everybody. An update from Lonap in the UK.

We are an association-based exchange for those of you who haven't met us before. That means instead of just buying services from a commercial entity, you become a member of our organisation. That gives you a great chance to have a say in how the exchange operates and grows. We are. We have all-time peaks of 8 gig a second, we are victims of the great weather outside this week. We have four sites in London Docklands where you are able to connect and long lining is acceptable into the exchange. We have a well had a matured Cisco 65 platform. 80 current members. 30 percent of or of which exclusively peer at our exchange. You are able to take a gig, 10 [unclear] and 100 [unclear] platform. If you are one of others members connected to a hundred meg port, then please organise a free upgrade to gig.

The other notable piece of news for this year so far is that we celebrated our 10th birthday earlier this year and this is a splendid celebration at the former site of the Bank of England.

We have eight new networks connected this year and four others in application. Application can mean various things, although in most cases it means that we are waiting for cabling. The makeup of our new members this year has been six British networks, a French network, a German network and four networks visiting or connecting from the [unclear].

We have a number of technical projects that we are working on at the moment. Firstly, we'd like to introduce a multi laterally peering service. At the last Euro IX meeting a number exchanges presented on how they did it and some of the lessons that they learned and one of my plans, instead of going and learning all these lessons from scratch again, is to meet with the guys who have learned those lessons, produce some best practice documentation and then share that suggested best practice with the other association based exchanges in Europe. So, that's how we are going to proceed.

Next week, our peering LAN is going to be dual stack. At the moment if you want to peer v6 and v4 you need multiple ports or you need multiple tag V LANs on one port. Our members have asked us to change that, so that you can peer v4 and v6 on the same port and this is led to a number of new networks saying, we'd like to peer v6 at the exchange for the first time so we are quite excited for the about that.

We are evaluating several new sites in the London region and if anyone in the room would like to talk to us about the new sites in the London region F we can assist anyone in the RIPE community or in our membership community to let you know what's happening in the region, come and talk to us.

There are two representatives of the exchange in the room and we like to talk and chat and dine and drink. So these are our photographs. Come and find us and talk to us. We'd like that very much and there is still a couple of minutes left if anyone does have any questions.

AUDIENCE: Alex. What's the percentage of exclusive ASNs in the low net?

ANDY DAVIDSON: Well, although I wouldn't like to give undue advantage to anyone who wasn't paying attention, I put that slide on the screen now, so anyone in the room who wants to win some swag

CHAIR: You can always pick up the presentations off ROSIE, so you don't need to ask a question.

Actually, not all the questions can be done off agendas I have been told. Next up is Mike and then Daniel from NaMeX.

MIKE HUGHES: So, good morning everybody. I am Mike Hughes, I am the CCO of LINX. I am I have go couple of minutes of what's going on with us apart from anything else that's ongoing that you can read about. Some statistics. Current peak 245 gig and that's a quick illustration what have a good weekend can do for your stats. We have our first good weekend of the year on the 26th April. And we were about 30 gig down [unclear] what we'd expect to see as our peak on the Saturday. So it's proof that people do actually stop playing games and downloading tore events and things like that and go out and have real lives once it gets nice. And then the little dip further on is what Foundary entries can do for your [unclear] as well.

Port friends: We have got 10710 gig edge ports connected. That's the light blue line in the bottom right of the graph that's climbing and what you can see is you can also see a corresponding fall now in the 1 gig and 100 meg. ports as those are being replaced by the 10 gig connections.

So, the total ports connected. Just shy of 600. The light blue area is the other graph turned into a stacked format and you can see again 1 gigports falling away and 10 gigports taking over there.

One thing that's having of course [unclear] only see gig ports; you see this is a port and whether you see whether it's a 10 gigport on a one gigport and we only count it once. What happens if you count the traffic you get a graph that looks like this that's going up to the right in a big way. So, the light blue gig is now this enormous light blue here, [unclear] 10 terabits of capacity. That's ports that have been sold to links members. We are not counting the backbone ports in this. There is another terabit's worth in there and there is no unsold ports in here. These 10 gigports make up the fast amount of the traffic. We'll see that's probably the same for most other exchanges as well when they start doing 10 gig.

We have relaunched our Sflow portal and it's totally reimplemented. It's based on an open source code at the back end and then our own proprietary code to give the front end presentation. We have got membertomember stats available. We have got engineering tools to help us [unclear] with the challenges of 100 gig not going to be with us for another two years. You can see stuff at the URL below. We have got future releases on the road map. We have been taking feedback from the members and working out how we are going to build that if the future, as far as this goes.

Again, it's open source tool but we are not open source in the proprietary bit of code. People ask where can I download this? Is it freely available? And so much of the bespoke code is link specific. We can't do that at this stage. The open stuff is he back end is [unclear].

The existing network is being expanded in [unclear] it of the space work we take in Telehouse North. We have got so much space and so many connections there. We need more space. We made no Foundary MLX 32 [unclear]. That increases our 10 gig capacity, gives /URS interswitch capacity that's bigger than 8 by 10 gig capacity. We will retain ring architecture and we are using DWDM switches, so we are not using out board DWDM gear, that gives us significant cost saving.

The other thing that touches on what Eric was talking about from Equinix, we are looking at new locations. Hopefully ready for service late quarter 3, early quarter 4. Discussions are very mature. We are basically finalising details for this. The existing docklands sites are in yellow on the righthand side clustered together. As you go off to the lefthand side you can see the new green pins and they are the new sites that we are looking at, which is interaction in the city of London in the financial district, tele city which is in the west and then way over near Heathrow, Equinix LD4, that's in the wonderful city of Slough where the UK version of The Office was filmed.

And that is it. If you have got any questions, you could ask me if you are feeling brave or you can ask one of these other four fine people that are with us in the back of the room. Thanks. Any questions?

CHAIR: Thank you.



DANIELE ARENA: Hi there, I am going to give a brief update on NaMeX, which is an Internet exchange point in Rome, Italy. That's going to be brief but it's going to contain one of the answers of your questionnaires, so...

It's always hard to come after LINX. We only have 26 members and we just had our first 10 gig connections now. We passed the 10 gig barrier, but to be fair, it was just a onetime episode and we are running mostly on the [unclear] 9 gigs.

What news we have. We have lowered our fees which were quite high, but now are more in line with other European IXP's prices. We have been doing some research activities, mostly related to our VoIP peering and multicast peering projects. That's been done in collaboration with National Research Centre, at least an institute of the National Research Centre. We are collaborating and now this is the answer you are looking for, with the department of Roma Tres University. Most of you probably know them from BGPlay, we have been developing IBGPlay which is kind of well they have presented it at RIPE 55, which is last time. It's a kind of the reverse of BGPlay, so instead of showing your AS from the RIS points of view, it shows interesting points of the network from your AS. So, what's cool is that it's all graphical, it shows the updates and well, it's interesting, you can check out the presentation on website and there is a link to their website, so if you want more info, you can ask me or them.

And it is hosted at our exchange point as a service to member ISPs and there are quite a few already involved.

But what's been mostly keeping us busy recently is the community. We have been doing quite a few meetings. Now we are at the fourth general NaMeX meeting dealing with technology and strategy and we have been doing the third peering workshop, but mostly what's going to happen in three weeks is the first Italian peering forum and the cool thing is that it is organised together by the four larger Italian IXPs, but this deserving a slide on its own.

It's going to happen in Pisa, home of the Leaning Tower, on the 27th May, 2008. As I said, it's going to be organised by four Italian IXPs, mix, NaMeX, particulars and topics. The meet something going to be both in Italian and English and translation is provided, so not only Italian IXPs are invited. Actually I am inviting you to come. You can register at if you plan to come. The goal is to create an Italian network operators group, as it seems to be fashionable recently, and that's it. Come to Pisa if you want to visit a futureful city. That's all. In time. Thank you.

CHAIR: Any questions? Okay. Remco and then Kurtis.

REMCO VAN MOOK: Good morning everybody, I'll try to be brief again. This is the [unclear] update for RIPE 56. We have got a new logo. Let me repeat that. We have got a new logo. I hope you like it if you don't like it, tough. This has been based on broad market research within our community and there were a few alternatives and there was one alternative that the Dutch side of the market liked and there was one alternative that the German part of the market liked and in the end the Germans won. This pretty much like soccer, isn't it.

Finally, some charts. These [unclear] have been coming to me for a while, why no charts, why no stats, why no figures? This is actually a bit different, so I am not doing traffic, I am just doing members and VLANs. As you can see, this is has grown quite significantly, I have taken my old presentation I have done here and carted them. So you can see that the amount of members has gone up from something like 60 to 350 over the last five meetings and the amount of VLANs that we are doing has gone up from 25 to 250 and this is ever increasing. This also makes for a rather complex infrastructure, as you can ma'am.

So, new network, we finally have a decision on what we are going to do. We are going to make a full core edge abstraction. For now we are going to do a he have ugly hack called local VPLS, if you are interested in the gory details, contact me off line and eventually we'll try to move to Mac inMac. These are the sites that [NDIX] is currently operational in. As you can see it's quite a large part of the middle of the Netherlands and some parts of Germany and there is some more sites planned and are not only in this picture, but they are in The Netherlands and Germany.

That's pretty much it. Any questions?

CHAIR: Thank you Remco.


NURANI NIMPUNO: Hello, I am not Kurtis, it's not my first presentation in English, but please be gentle anyway.

So, Netnod, we are located in Sweden, we have got six exchange points in five different cities. And we have actually reached above 50 participants at the exchange. So that's a great step forward for us. As you might know, we are located in these underground bunkers. I got a question from someone who thought our offices were located in underground bunkers, but no, we actually sit above ground but our equipment is located in underground bunkers. So that's sort of what that looks like for those of you who haven't seen that.

So, we also have a set of common infrastructure service that is we provide at the exchange, official Swedish time through NTP, we also run I-root servers dot net and provide telecast and a number of other services that are available.

So, well the situation in Stockholm is a little bit special also because we are located in these bunkers and in Stockholm because the location of the bunkers are not disclosed, the charge include the cost of the fibre to the bunkers, so that's why, if you look at the fee schedule for the Stockholm exchange, it looks it's slightly more expensive or quite significantly more expensive.

So this is what the pricing structure is and we went through, in 2007, we further reduced the fees and we actually went through another round of that in 2008, so that's what it cost to connect to our exchange in the various locations in Sweden.

We see quite a high amount of traffic per peer at Netnod, so although we have got 51 peers at the exchange points. We have around 2, 2 and a half gig of traffic per peer, and these are the top 20. So that's quite positive to see obviously you want people to peer if you connect to an exchange point.

So, this is the cumulative traffic. So this is looking at all the five exchanges in Sweden. So we reached 100 gig in 2007. We have seen quite a stable growth since then. 140 in March, and that was just before the weather turned sunny and warm, so after that, it's decreased but I can assure you that the weather will turn bad in Sweden again, so you will see an increase there.

So, this is the traffic in Stockholm where obviously most of the traffic takes place.

This is my ability to draw mountains over time. So you can see quite a clear improvement there and I think I have reached yeah, a good quality there at the end.

So, some observations. Like I said, the traffic is still increasing. We have reached membership now over 51 I think, I don't know if we have got a few more members in the last few days. Actually, I think at the last RIPE meeting we were around 41, so we have been pretty busy at the exchange for the last half-year or so. We have seen an increased interest from a few countries as well, including Russia and some of the other countries around that region, so that's interesting. And we can also see a continued demand in 10 gigports, so we get new members connecting with 1 gigports and quite a few of the existing members going up to 10 gigports.

We have upgraded our router infrastructure over the last year to support the 10 gig. We are also looking into VDM or WDM option to say load the cost connecting in Stockholm. So by actually putting WDM equipment at the data centres we can sort of cut the cost of the fibre for people connecting to the exchange. We have got IPv6 enabled and I think Curtis will talk a little bit more about that and we are looking at more, we have organised an IPv6 workshop and we are looking at organising a bit more customer events and workshops and thing like that.

And we organised the social yesterday, which, I unfortunately couldn't attend but I hear [unclear] a good one, so...

And that was it from me.

CHAIR: Any questions? Thank you.


JOSEF CHOMYN: Hello everybody, I am Chairman of Board of Neutral Internet Exchange in the secretary public based in Prague. I'd like to apologise in engine LIR but for sure this is my last presentation. I will explain it later. You can see that we have new logo on our sides. But I don't want to confuse you with filling the questionnaire, because... it's nice.

So what's new at our exchange point? We have more members, more networks. More ports. So, this shows some statistics that the number of 10 gigports increased to 20. The number of 10 or 100 ports has decreased, that's obvious. And the current numbers, the current number of networks is 77, plus 4 TLDs, that means totally 120 free ports. We have some IPv6 networks doing peering, and for the traffic, we have reached 40 gig peak last month, that's compared to 30 gig in October. So it's increasing quite rapidly. As I said, more technical information and much more nicer presentation will be next time given by this man who is a new technical director and that's because we are going to work for professional organisations, so it's fulltime employee and it will be much more better than me. So thank you.

If you have any questions.


AKIO SUGENO: Good afternoon, I'd like to give you an update of the Telehouse exchange point.

New York NY IX statistics is peak is 44 gigs, average 33 gigs and the customer number is 104. In the last year we upgrade our SH to Foundry RX platform and again available locations at New York is 25 Broadway, 60 Hudson Street and 111, 8th Avenue.

In response to the demand from the customers, we are plying to extend NY II X to 7 teleport drive by the year end. Trying to explain the geographic locations. 25 Broadway, 60 Hudson are all here and the other is located in Staten Island.

The wide peering at 7 teleport because there are two reasons. We still have plenty of power available at is [unclear] set teleport after two phase of power upgrade work. Current capacity is 8 kW /rack and it's still upgradeful and the third phase will be starting early 2009 to double this capacity. Also the connectivity is pretty important. People realise that we need better fibring structures so they started building their brand new fibring structure in New York area, what we call new fibre goes to the, goes to a much larger ring than the conventional fibre ring, so old fibre terminated in 7 teleport, just the reasons we have a lot of customers start to go migrate to 7 teleport.

Also, we have a lot of the partnership in NYIIX is available at one of these locations, avenue of America or [unclear].

LA peak is 5 gig, number of customer is 46 and we ungrade the switch to Foundary RX early this year. In the available location is 626 which will share and 1 which will share.

Route servers: At the moment we have over 10 /PUPB customers, we have started to you know, install the route serve [unclear] at both locations in New York and LA. This is [unclear] and we want to make sure we separate MD5 for the IPv4 and IPv6 both protocol. And we are scheduled for launch by the end of the second quarter.

And lastly but not least, when I visited the DECIX, a DECIX team member shared the experience route server DECIX with us and we learned a lot from them. So I'd like to take this opportunity to thank DECIX team for their kindness and support.

I think that's it. If you have any questions, please give me, send me an email or grab me, I should be here for today. Thank you very much.


CHRISTIAN PANIGL: Hello, good morning, I am working at the University of Vienna, we are operating the Internet exchange. I am giving you a very brief update today. There has been a more elaborate update at the last RIPE meeting, if you are interested in what we have been doing for the last twelve months, please look it [unclear] at RIPE 55 repository. During the last six months, we have been concentrating on improving port security features and thinking about legal constructions for allowing remote peering.

Therefore, I am reducing my update on the policy update today. Aiming at an increasing number participants, we have doing some changes from, and those changes are now valid from May, 2008. So from now.

We have been relaxing the prerequisites for a v6 connection which means that there is no requirement to be an ISP to be connected to the Internet exchange but we allow basically every AS number holder with a properly established international Internet connectivity to also connect to the Vienna Internet Exchange. (VIX) and what has been announced for a couple of RIPE meetings now, we are now ready to allow remote and distant peerings base on the improved port security and improved monitoring tools. These improvements have been basically made, or primarily made by my colleague, Wolfgang, who has also implemented the Sflow statistics and Sflow peering matrix and other things, nice tools on the website and specifically insight the web portal which is available to participants only.

And we have been lowering the port costs generally, and specifically aiming at enhancing the connection quality and redundancy. We are going to lower prices additionally for dual site connections. We have different sites in geen I, one is at the university itself an one is across the river Dan you'll in the north of Vienna and those two are redundant [unclear] interconnected and therefore to improve the overall redundancy for the Austrian, specifically for the Austrian Internet infrastructure, we are now offering 30 percent discount for dual site connections for those who are really connecting to both sites at the same speed and doing low chairing on both locations.

That's it. Because it's RIPE, I am not allowed to tell you any prices, but Vienna is only talking about soccer these days, so it's dedicated to the soccer Champions League, we are lowering the prices, look them up at our web page and who is interested? I have brought some brochures, they are fresh, so please help yourselves. Thanks so far. Any questions?

GAURAB RAJ UPADHAYA: Thanks Fergus, I'll just do the Asia Pacific IXP update. I did this two years ago in Istanbul and I lot of people came back so I'll just do a quick run down.

I am going to like the where everybody needs to go or, definitely Japan and when I say greater than 10 gig it's both in /ERPLS it of traffic as well as in terms ports. Most of the access points in Japan seem to have you know, fully established 10 gig platforms and so on. In Japan rather than saying there is a single IXP, there is the automatic building which seems to have the all the different /WEUFPS and probably all of them do quite a big volume of traffic.

Other important Internet points should be Hong Kong. It's almost all current connected to, it's also run by the university and I think a bit lacking in innovation and I think that's why Equinix is setting up their own platform in Hong Kong.

HKIK does offer 10 gigport. They are currently looking for 100 percent or doing a platform upgrade to have, to be able to provide more 10 gigports. Fully they'll be done this year.

The other big volume traffic exchange in Asia Pacific, Korea where it looks like the switch is no longer used, and all the big providers actually have cross connects, so, it doesn't matter what Corrie there, but the switch there if you consider that all the cross connects between the ISPs are exchange points, probably the biggest traffic exchange in the world in terms of like hundreds of gigs, but no switch.

In Taiwan, there seems to be still be a fragmented IXP market with [unclear] of transit providers running exchange points and people running their own interconnects and so on. Again, not a single unlike Japan I don't think there is a single building in Taiwan that has all the different switches.

So so those are the big ones. The medium ones, the most significant are the important one is Singapore. Singapore continues to be like, when I did this presentation last time, in Asia Pacific, if you are going there it's probably going to look at Tokyo, Hong Kong and Singapore as points to set up exchange points, or points to set up a POP, so Singapore continues to be more and more important and in the last two years I think specifically Equinix has got providers from Philippines and Indonesia and all the other regional players, but I don't think they are still doing 10 gigports in any exchange points in Singapore. I may be mistaken on that but last I knew there was nothing like that.

But all of them are looking at 10 gigports.

Then on the smaller size, around 1 gig of traffic, mostly 10 to 155 AS P ports and the person Australian exchange, pipe networks, they do it on the peering fabric in Sydney and the rest of Australia. There is also Equinix Sydney, which is also getting some tracks on. It's a much easier market with established peers and mostly port and you probably don't expect new providers hauling all the with a I to Australia to go there and peer. Maybe see more content providers.

New Zealand, same thing, there is April, which seems to be you know, (AIP) more of an international great way to the Wellington exchange which holds a lot of traffic. There is a couple of others within New Zealand and but none of them are big or significant in any way.

NIXI continues to be a case to look for you can take that as a positive or negative. It's getting there recollect they recently did RFP switch upgrade which means that the capacity on the RL6500s and probably are looking to upgrading to 10 gig platform as well. So, it is getting at least seeing a lot of traffic now.

There are quite a few small exchange points. There is two new ones, there is one in Manila, the Philippines open exchange established last year. There is a new exchange point in Kuala Lumpur which I am not so sure if it is actually operating, but I will know more about it. Indonesia, IIX, seems to have, not? Jakarta but also in [unclear]. VNIX, mostly domestic traffic. Of course we have the NPIX if CAT man do you which add add second switch evenly. W D IX, in dak a, they are growing also, last year they had tomorrow issues there, but finally they are connected with C U ME 4 and you know the traffic there is growing.

Colo, there used to be [unclear] fragmented number of IXPs, [unclear] down to two different switches. If you have seen reference to His pan and IX in Colm owe, there is sort of combining to single NTT, with two switches.

Monday goal also concentrated back to the open mix. There is two exchange points and one was not able to up their capacity in time and switch and so on. So things have seemed to be calmed down in open mix.

All of these exchange points are probably less than a gig of peak traffic.

Of course, there is interesting situation with these other big couldn't consent and you don't see a lot of exchange points because mainly to do with transit, Monday op please or due op please, Singapore is a case there is the biggest provider and runs the socalled sticks, which is the exchange there and quite a bit of Asia Pacific connects through sing tell as a transit provider and they make things difficult enough for Equinix and socks that local capacity price something so expensive Singapore and that is probably one of the biggest problems for he can initial and /SO*BGS in Singapore. Though you can probably talk with the Equinix people here about what tricks and techniques they have devise indicate Singapore to make it [unclear] happen.

China, same thing, big providers, no real access point within the country. Thigh land continues to be the same thing. Thigh land did regulate and I think at least one of the providers called true got a circuit directly going to Hong Kong now, but still there is nothing Malaysia of course there is I can't figure out what's happening in the new exchange points that's been established there. Pakistan continues to be a monopoly for international transit and there is no exchange point. A lot of other countries, either they are too small or their fail regular tear that there is no exchange points in those countries. Thank you and if you are looking for more you know, probably you should come to the APRICOT in February 2008 in Manila, Philippines.

Thank you.

CHAIR: Since we are running ahead of schedule, which is nice, thank you for all being so concise, [unclear] is going to be the you're ex presentation before lunch rather than after lunch. If people would like to start passing the question papers forward, we can start counting all the answers and finding out who is going to win some of the many prices. It's probably going to be a prize for everybody actually, so...

SERGE RADOVCIC: I guess I'll keep with the running joke about the language thing. I am from Australia for one. And secondly, last night, during I don't know, one of the songs when everyone was going off their heads, someone head butted me from underneath while they happen to have my [unclear] sticking out of my mouth. So I have now got a huge hole through my tongue. So if you hear any sort of [unclear] link sounds while I am talking, you'll know what's going on, yeah, and it's coming out of my mouth.

So, anyway. For those of you who don't know. You're IX is the European Internet exchange association, as our name suggests. Not only in Europe, we have some members outside of Europe. We are not an Internet exchange point. We are not an Internet exchange point. So I still get males every week asking how much it cost to say connect and get a 1 [unclear]. We're not an Internet exchange point. We are an association of Internet exchange points. We are an association. So I think I have made that clear. I just helps with the emails.

So, currently we have 42 affiliated IXPs and 4 patrons, thee patrons financial support EuroIX and they join in the forums and mailing lists.

36 in Europe. Two new member IXPs, crow Asia has joined us and another from row main a, inter LAN IXP. And our very latest associated member to join is the [unclear] Internet exchange, NPIX. We are proud to now have four exchange points in Asia and our first [unclear]. So welcome.

So, I mean, there is 36 ISPs are pretty much the biggest ISPs in you're /OFPLT with a gain of these [unclear] we also LoST two from France in the last few months, so now there is no France isn't represented at EuroIX any more, so if you are in the room somewhere, you are also ducking and trying to run away from me but I'll probably try and catch up once again and see if you are interested in joining.

Member traffic: This is just the 36 that I am talking about. Over the last twelve months, it's gone up around 75 percent, which is sort of down on previous years. It's usually used to be between 80 and 90 percent growth rates per year. Since the last RIPE meeting, it's about a 250 gig increase. Our percentage is down, but I mean it's still big increases in traffic. At the start of the graph there, you see the normal summer drop in traffic and then the big drop later is the Christmas, the New Year's Eve period. But then for the first time, I have been watching this traffic for about six or /SEPB years, I have saw a drop in aggregated, 90 percent of the members actually had a drop in traffic around the end of February. A if you of IXes have this regularly do you to school holidays and other things but not across the entire membership. So that was unique. Since then traffic growth has been a little bit slow.

The green is the member IXPs at the bottom part of graph. The purpose he will are not member IXPs who publicly display the traffic on website and the top blue bit is something I have been getting since about the end of August, they are IXPs who don't publicly measure the traffic but they have given me these statistics which I have added. My best guess at total aggregated peak traffic in Europe at the moment is around 1745 [unclear] per second.

A little bit of a comparison with Asia and the US. Admittedly I don't have as much access to Asia and US sites as I do in Europe. I have contact with around 80 or more IXP /TPHS Europe so they are a lot more /ABG set but I think it does know a relatively true trend in Asia and the US. You can see that all three of the couldn't [unclear] had that drop around the Christmas period. Europe then shut up around the end of December to, through January. But then had very slow rate. In fact, in the last five months, it's only around 15 percent increase, which is probably the slowest five months we have seen for quite a long time in Europe. Asia, also a little bit slow. US seemed to have been the big [unclear] rates. Like, again, I don't want to say these are absolute trends because there are a lot more exchanges in Asia and US, but I think pretty true.

As far as participants go, I was a bit surprise that had Remco said he had 350 participants so I think I am missing quite a few from my numbers. These are the ones that I know of at least so it's this plus a little bit. I estimate around 4160 participants at the moment at European IXPs. Well, more than 600 of them peer at more than one IXP so this brings us down to like a unique ASN base of around 2204 across Europe. As I said more than a few peer at [unclear] will IXP. 17 peer more at 10. 17 of those peer at more than 15. One I know that peers at least 20 different European IXPs.

Newer anney was talking, you mentioned that the Swedish traffic was armed 2 [unclear] per participant F I do an average for Europe of that 4 thousand participants, it's around 420 meg.s. if I take that to the [unclear], it's almost double that around 790 [unclear] on average.

You can get hold of a lot of this information by going to your owe IX dot slash resources. There is a lit of European IXPs and a direct [unclear] to their traffic statistics. You can also get into the ASN database which has all those ASNs list and you can see which IXP they are at and other information. We also produce reports for the last few years on the European IXPs, you have also got access to these reports at the resource as part of the web page.

Since the Amsterdam meeting, RIPE 55, we had two Euro IX forums, 11th Vienna in November of last year and just two weeks ago in Stockholm. At boast of those events we had more than 36 IXPs from around the world talking about IXP specific topics. Technical, commercial, regulatory matters and a lot of socialising as well, not quite the same sort of the socialising, they were a lot better behaved at Euro IX forums. Our next forum is in Geneva on the 10th and 11th November. It's going to be hosted by the CIXP. If you are not a member of Euro IX, but you are an IXP, this is an IXP only event and our lovely patrons of course, please feel tree to contact me and you can see what I can do about inviting you to the forum as a prospect.

Well, on that point if you are not an IXP that's a member of Euro IX, please come and chat with me, maybe I can convince you that you should be a member of Euro IX. If you are an IXP participant and your IXP isn't a [unclear] Euro IX. You can convince them to have a chat with me or I can give you information. Go on our website, contact me. You can go to the website, and you will see a lot more information about how you can become a member of Euro IX.

I think that's it for the European update. Are there any questions?

Thank you.

AUDIENCE: I just wanted to know what a 10gig port at Euro IX costs?

CHAIR: Thank you for that helpful question. Thank you [unclear].

Not everybody apparently got a questionnaire, competition papers, if there is anybody who has got spare ones there is a group of people up there who want some swag, and one or two would help.

Anyway, we will mark your exam papers, and we won't highlight anybody's stupid answers but we might highlight some of the more fun ones we have seen and we will do a distribution at the beginning of the next session.

So, lunch time and we will see you all this afternoon hopefully. Thank you.


EIX Working Group Session 2 2:00pm, Thursday, 8 May 2008

CHAIR: Good afternoon everybody. Sorry for the slightly late start. The other cochairs and the describe were all work stuck in meeting discussing policy development process, we haven't finished you counting the competition stuff we will do that at the end and.

So, moving on, the next thing is Greg talking about IE and 1,400 Gig.

GREG HANKINS: Good afternoon everyone. Here is the standard disclaimer slide that are you used to, I don't speak for the IEEE user. These are my views and I am no, no way representing them in an official compass I there are some slides that I put in here because there have been some new and exciting developments. So basically, last year, [unclear] not going well around this time in May, we had kind of a dreadful meeting in the basement of the ITU building in Geneva and in July a lot of compromises worked out and things started moving forward. We are going to do 40 and hundred which are like things you need to get out of the study group state machine and then [unclear] approved by the Standards Boards in December so we are moving right along. We are now the 002.3 BA task force, you get your letters so.3 BA is is a 48100 Gig. One standard that will be specified for both speeds so this means that they are coming together. So I always make this point for those of you thought 40 Gig would come before 100, that is not the case the standards will be delivered together and the big underlying theme was to reuse a lot of 10 Gig technology as well as some copper and optical, [unclear] a stuff already out there. Here is is a summary slide: Basically, doing everything that ethernet does already. And this is the objectives and you have seen this slide before. There is one objective that was just added in March, it is the 40 Gig 10km, so before the application for 40 Gig was targeted at the data centre and computing environments and then some interesting things slipped in like compatibility with OTN and then like a SIG mode fibre reach for 10KMS so it's a competing technology for 100 Gig and we will see what this does to the market.

So the architecture options are under consideration. Now two are being talked about how do we do 40 and 100, so do we do a multiLAN solution or [unclear] I can't solution. The MLD proposal is the winning one right now, PPL which is option number two, I think pretty much only is driving that one and there doesn't seem to be a whole lot of interest and the gory details you can download this presentation by Larry Green and read all about it.

So the one vague specification is to provide appropriate support for OTN. And that is all [unclear] says and there is no details and criteria about how to could that. So basically, things are going well, there is an official liaison and a couple of people from the IT IEEE also attend the ITU, so the ITU has voted to define a new tier for 100 Gig and we are not sure how this is going to be done yet T depends on how the 100 Gig specify case is written and second of all, kind of what options are available and then for 40 Gig, there is going to be some way of mapping it into ODU 3. How that is going to be done is still under investigation, and it will be some sort of transcoding option, we take some bits and then turn in some other bits and add some stuff to.

As far as the objectives, there is a lot of progress right now, will there are some proposals as far as technical feasibility on thousand do all these things that we have decided to do, and basically back playing ethernet, this doesn't interest a lot of people but there are a lot of [unclear] by the server vendors, connect all the components or something like that so basically, the current proposal is just to do 10G based KR and turn that into 40 G based KR4. As far as copper cable assemblies, the current proposals are also to take the 10 base KR technologies and just over copper. Copper is /STRG for fibre there are well known and standard specifications, OM 3 multimode for example. For copper there is a lot more freedom because there is a lot of room to define how this is going to look so it could be a twin axe it, could have quad S F, there are proposals on the table to investigate all the technical and economic feasibility and choose some sort of solution so there is no standard copper cabling specification that is going to be able to support these higher speeds. It's not going to run over category five, for example.

So, for 100 metres OM 3, there is really three things that people are thinking about: The one that is most popular is a multimode parallel fibre. This is has been used since early 2000s and it's there is obviously deployed technology and that is always a big win.

There have been a couple of discussions to extend the reach beyond 100 metres or possibly to do something with OM 4. So there is a cabling vend they are a now, starting to talk about OM 4 and the IEEE, everyone has their own little agenda as they do in all standards bodies.

And then the big discussion is how to actually do 100 Gig ethernet over ten kilometres for 40 Gig ethernet, that was just adopted in March, so the consensus it will probably [unclear] CWWM grid. For 100 Gig 40, the consensus is to use LAN WDM and for ten this is where all the discussion is going, it's /AP key area we have to figure out next week, the IEEE is next week in Munich, and there are two major optics vendors which is [unclear] and OPNIX which is pushing their solutions and equipment vendors used optics and then customers used equipment so there is this whole like higher key and politics and it's a kind of complicated.

But basically, this is one of the last remaining big disagreement issues and there were presentations and discussions and motions made in March but we know we have to figure it out next week. Other questions that are being answered: There is a lot of block diagrams and terminology that has to be defined, CGMI and PCS and MAC and PMA and D all these things actually to people that design A 6 and make bits move, it's important so, maybe not to us users but this actually does have to be defined and this is the 100 Gig architecture that is being processed and thousand would all stick together. One of the key things to point here, in the initial implementations will probably be a multiplex wave solution, for 100 Gig 4 by 25 but we know that several years when the optics neck following carries up we can do 100 Gig serial, all of these architectures are written so it can do multiplex solution and then a single wave solution when that technology is available or feasible.

There is lots of management things that we have to define, so MIBS all this kind of stuff and one that people don't think about ask actually there have been test procedures to make sure that all the bits that come in go out, and the exact same order.

And there are test patterns that have to be written to make sure that that happens.

This is a slide I started to put in just to give you an idea of how big this project is. This is like the kitchen sink project of Ethernet so. If you look at 10 Gig, there are a number of [unclear] have been developed over the past five years, for 40 and 100, number one we are doing two speeds at the same time and number two, we are doing every imaginable reach and medium that you can think of. So it's a really, really large project. And maybe that is why it's taking a little bit longer to finish.

Here is the where we are now slide. This is the short version which I usually talk about, I am going to skip to the long version and talk about this one instead, because it's actually starting to get interesting and we are about two-and-a-half or three years into this whole process and what everyone is driving towards is the D 1.0 date. So if you look over there, it's in September of this year, draft one zero is what search waiting for because when it's written then people can start to go off and implement things with reasonable certainty that it will work as a standard when that comes out in a couple of years. So September, all the vendors are looking towards that date, when draft 1.0 is out we can start building stuff, component vendors can start and it's off to the races then [unclear] future meetings, like I said, the May meeting is next week in Munich and I start today put some key things we are doing. Munich is really a key date because this is the last new technical proposal that can be accepted so we have to figure out how we are going to do 100 Gig, we have to figure out the copper cable options and things like that, and then after May, all the options have to be on the table. So there can't be any more objectives or any more proposals or any more here is how we could do this, type presentations. In July we start making choices, selecting what is going to happen, looking at all the proposals and then choosing, well the one that is most popular, I didn't say the best one, the one that is most popular, and then in September, we start writing 1.0, and that is the big date that everyone is working towards.

More information is here. This is the public area for the, all the presentations are here, there is a mailing list you can join and tons and tons of fun stuff to read and then these slides all skip but this is the very detailed IEEE standards process, you can read through that in your spare time.

OK. And that is it. Any questions?


ARIEN VIJN: OK. This is about the same presentation I gave yesterday in IPv6 working group, but this meeting is more towards IXP so we might have some discussion. Please interrupt me during my talk if you see the same thing or have some concerns.

It's about issues we see on the M6 related top IPv6, and as I explained yesterday, we are a layer 2 internet exchange so what the hell do we have to do with IPv6 for us it's just payload. Well it has to do with the multicustom and flow of traffic we see coming out of the LAN, just by hooking up monitoring system into the v. LAN that just looks at all the traffic that is being flooded to that port. It's also about the port security issues we see. It's many Internet exchanges we only allow one, if we see another we just block or shut port. And once that is done, this sends a signal message to our sys lock collector and, yeah, what do we see then? Well, that is the first topic. We see a little sys log and I don't expect you to read this but we noticed there are SYS violations that are fairly typical. They look an all of lot like the IPv6 prefix we use on the peering LAN 2001 0 F 8. Column one is our allocation. Sys log of course which port these violations are coming so we could correlate a few things. We can look at the OUI of the source that is allowed. So we can see that this typical violations are only coming from Juniper routers, and what we also notice is that they are coming from all types of interfaces, 100, 1 Gig and 10 gigabit, and they are only coming from v6 enabled Juniper routers.

We could also see that these violations come during the time, A, when random v6 peer is not available, because their router has crashed or is not reachable. So we did some research and captured some of these violating frames what we saw is they are encapsulated BGP messages. We informed Juniper about this from people working there and they actually tried to reproduce this in their lab and they were not successful and what I hear from some other Internet exchanges, but not all, is that, yeah, other Internet exchanges don't see this particular issue so it's kind of strange. Is it only bound to M 6 or are other internet exchanges actually looking to this, can correlate this kind of violations? Is my question, actually.

Although it doesn't seem to be harmful, this particular issue, we may shut the port if we see Juniper is pretty good in not doing wrong things, so members actually don't have any problems with this particular issue. Only we see it in our sys logs a lot.

Another issue we see with IPv6 and on the internet exchange is that we see bursts of ICMP enabled discovery queries, who has this particular IPv6 address, but the IPv6 source of these queries are actually not the address that is bound this bound to this particular router and it's our only three out of nine routers doing this actually it should be Foundry there, IPv6 enabled Foundry routers T doesn't seem to be harmful, I was wondering why is it not harmful. Well first of all the addresses are kind of bogus, all zeros, colon colon, prefix and then the rest is zeros. It's addresses that are from within the AS of the particular BGP router, or local link router with the last anybody he will in network order wrong. So it's a whole variety of violating, but it is also about source addresses from other members. Although that is quite rare. It's really the vast minority of all the queries. And I was wondering why is that not harmful. Well, in most cases, there is a correct query directly after it, so things are corrected again directly after this issue. Nobody seems to notice and nobody seems to have problems with it.

Then, the last issue we see relates to IPv6 is more like a housekeeping issue; it's about ICMP multicast listener reports, lots of them bursting out on the and yeah, there are of course response on ICMP listener queries and that seems to be a new feature in CISCO, people enable IPv6 multicast routing /WB within their own IS and the solution is for CISCO people to no IPv6 MLD router in the Internet in interface context.

So the conclusion is just like yesterday, it's all mostly harmless, all works. Juniper specific problem might cause a security port security violation. But yeah, that rarely happens. We have the Foundry issue under investigation. I guess we will inform the members involved and also contact Foundry themselves and the CISCO million pounds messages are a housekeeping issue and we will update config guide on the website so people can add to this to their router configurations.

So that concludes my talk. I guess there are no questions?


CHAIR: And then Greg again. It's me.

GREG HANKINS: It's me again. That is joint presentation and project that we did together [unclear] who is in the audience and Tom shoal who is not in the audience. We got together and said we should do a peering survey, and we did.

So first slide with some results about the results. I was pretty happy about the results because we had a huge number of responses, thanks in part to a lot of people that were asking people to take the survey and reminding people so. 193 started, we had some packet lost and ended up with 1534, that manage today answer all 24 questions. I think people just got bored clicking through the pages. We had 74 people provide their ASN, that was optional, we made the whole survey anonymous to avoid legal or whatever kind of issues, and like I said thanks to everyone who promote it had or participated in the survey.

So what did we ask. We had some general peering questions, some questions specific to BGP and then some other peering practices questions. The first question was what type of network do you run and I was surprised to see how many content providers are out there.

So IPv4 unicast peering. Yes this was actually a serious question in, some of the comments people made it clear that they didn't understand this was a serious question but I thought there could be someone out here building a v6 only network and not do any IPv4 peering at all, I don't know if that is good idea but feasible. So there is definitely a trend to deploy IPv6 this year and I think we had something about IPv6 at the conference earlier in the week, too. Basically, though, it seems to be still a chicken and egg kind of problem. A lot of comments were saying there is not sufficient demand yet, some of the also answered their application doesn't support IPv6. But, a lot of people are investigating and testing IPv6 this year. There have been a lot of discussions at sic IX and after SIG IX and APRICOT on the list, a discussion about how an IX should provide IPv6 service and most people prefer to dual stack and you can even see in the IX updates this morning and over the last few days there is a lot of interest in supporting v6 peering. Multicast peering, not a lot going on. A lot of no plans answers.

Again sort of the same comments as IPv6 that there is not enough demand or my application doesn't support multicast. Some people were testing but you know, the overwhelming majority of people were have no plans to deploy v6 multicast peering.

BFD, this was a good one; there was some concerns about stability and interruptability but it has been a hot topic as well in the community. There was an operators BoF in San Jose [unclear], some good discussion there and I have been talking to a number of IXes and IX updates and people are thinking about BFG, there are some presentations in the past couple of years about BFD. The kind of overwhelming theme again, though, is a lot of people don't know what it is so there is again opportunity for education here.

BGP communities seems like a lot of people use communities and MEDS, these are boring questions and I am not sure how to make it interesting. Path attributes, this was a question we asked basically do you change BGP attributes, a lot said yes and the overwhelming comment was that people provide a way for customers to do this automatically with community so is they can use ASP path pre pence or black holes or something like that. This was a question that actually also came up at the NaMeX peering BoF, change to a network you don't peer with. 11 people said yes, the consensus and discussion was it's hard to detect and Juniper CISCO have is a feature to mitigate this but there is not a whole lot you can do about this.

MD 5 signatures, this is again a big topic of discussion, people generally hate it and say it cause more problems than it solves. There is a lot of interoperability feature support and then lost password and kind of issues and we don't know what it is any more. One of the most interesting comments was to knee to add a sigh, let's do if you really insist. Question. We need something that examines the pros and cons and maybe best common practices for MD 5 it doesn't seem like anything is out there that talks about this, really.

The GTSM, so [unclear] and Tom did a survey to ask about people's plans and where this feature was going, and then we asked it again on the survey this year, and there was kind of course correlations, again people don't know what this is or they are not sure if their gear supports it and not many people have plans to deploy it. But many would like to use this feature, so I took this opportunity to sort of put this circle in there that talks about customer and vendor support and for us we are a small company and we prioritise features based on what our customers ask us for. We have this are you doing this and we say well, what are your plans and they say are you doing this. And we say, well, what are your plans? And so then we get into this circle. And it seems like if customers all start asking their vendors for this feature because we do believe it is important, then that will move things along and help us prioritise, so please ask us for this feature.

Four bite ASNs, waiting on vendor support and customer demand. I know kiss foe force 10 and Juniper will support it widely. I know there are a number of CISCO implementations for various platforms support it today and make it more available on all platforms. Quagga and open BGP also support T someone has customer using it. I was excited to see that and there seems to be a lot of interest in this one and I think that is very real feature that people will need in the future.

The MTU was an interesting discussion, many would like to use a larger MTU but they can't really because everyone has to use the same MTU on the peering fabric so Netnod does something interesting, they have two peering LANS one with standard MTU and one with 4470, is that right? So that is one way to get around this. But it's clear that we need to think about this and that it's hard to do so Tom also had a presentation on [unclear] packets at the Internet and kind of goes into this in more detail.

So what was your biggest concern about de-employing new features? This was also the most confusing question on the survey, and even trying to figure out what the results were was also confusing because I had no idea how to use this question thing that survey monky uses so I kind of tried to do the best could I with the results, and basically, there is a lot of concern about stability and interoperability and most of all being forced to upgrade code. We had ask discussion about this at the EPF so the themes correlate here and this may be in answer to why people aren't using the latest and greatest features. And something that we should have asked, which became apparent afterwards, was what geographic region are you from. So it would have been interesting to see trans sort of a region basis, so maybe we can put that in the next one. And then some conclusions: So, it looks like we are doing more IPv6 peering. Hank sent me this graph where basically, you know, if you draw a lines you can could you pretend there is an increase in IPv6 traffic, and then Google also has their IPv6 initiative now. So it seems like more people are talking about and the IPv6 stuff at NANOG and at RIPE this year.

Not much going on with multicast peering. There is /TKAOEF interesting /TR* in BFD, GTSM four bite ASNs and larger MTUs and the overwhelming theme and this is actually not a bad thing, a lot of answers on the questions were "I don't know what this feature is." So it's nod a bad thing if you answer that way or if you answered that way. It just means there is an opportunity for education so I think there is a lot of interesting features out there that have been written into code and are supported on major router vendors. So it seems like we need to talk about this more and education people. Tom is working on something nor NANOG 43 and I was kicking around idea with Philip Smith on some sort of workshop and we need to pick up that discussion again, wherever Philip is.

OK. Here are the complete survey results, it's actually pretty interesting if you want to look at all the questions. I didn't put all the questions in here, some of them out not to be particularly interesting. So that is the end.

Questions or answers or comments, observations?

AUDIENCE: Question, follow up on my last question to you at the GPF meeting: Did you do any more work on finding out if that EBGP multihop slide is correct or bogus as I suspect.

GREG HANKINS: Yes. So the question was about we had a question on there which I took the slide out of this presentation but it [unclear] asks how you prefer to do load balancing, and the options were link aggregation, EBGP multihop, multipath and a couple of others, and people overwhelmingly chose EBGP multihop I think. And the conclusion, then we had some discussion and the conclusion that was people didn't actually understand the question, because that doesn't seem like a very reasonable /TPAEPB you think about it. So that is why I took the slide out (answer) which again points back to I think, you know, there is a lot of opportunity for education on features and things like that.

AUDIENCE: Nick Hilliard, INEX. We did a little bit of stuff with Jim bow peer LAN at INEX and what we found out was that, well, people are actually quite interested in it until they realise the can of worms that it opens up in terms of raising MTUs universally on their networks because once you look into that you will suddenly find you are black holing traffic all over the place and because most end user boxes don't actually use anything greater than MTU 1,500, debugging this stuff and monitoring this stuff is a complete and utter pain in the as our jumbo LAN is still there there is no one connected to it but it was an interesting experiment.

AUDIENCE: I had also comment on the jumbo LAN issue because a few years ago we ask our members about jumble frames and basically the /AEFPBS, well, our customers and is either high providers 1,500 bites frame size /SOZ it would be only a problem to have a Jim /PWO*PB.

CHAIR: Any more questions?

GREG HANKINS: The final comment that I want to make at the GPF when we first present this presentation, we asked if we should do this sort of an annual basis and a lot of people there thought that was a good idea so I want to ask the same question, do you think that is good idea, should we do this on an annual basis. Are you awake?

CHAIR: Do we have consensus? Does anybody think we shouldn't do it? Sounds like consensus. Thank you, Greg.


CHAIR: Next up is Mike. Kurtis still seems to be lost OK, he is not lost in action. We have the winners and do a prize giving and get a presentation sorted out.

MIKE HUGHES: So first of all apologies to Richard O'Brien. You will not be seeing me in any fish net tights or scantly clad clothes. If you were the [unclear] Fort Lauderdale a few weeks you will seen this. It is environmentally friendly it is recycled.

So, quick history of time. LINX have been offering a time service for a number of years. Before we had the original stratum one time service these datum time servers we had a UNIX box it was like a [unclear] stratum two box timed against some other timed sources and near 2000 came along and London is on zero degree line, green itch meantime and everything, and that runs right through the middle of telehouse where our stuff is so there was this kind of marketing thing happened and this kit was donated to us so there was these three boxs in total and they were consisted of two devices, there was the cesium standard itself which was just like the cesium beam tube inside to provide the reference block clock and there was GPS to basically train that and get the time signal. That product that we have got they stopped making it and they are going to end of support it very shortly. I didn't have any photographs to hand but as Google is great and as is Photopick and we found this picture. That is the cluster in telehouse and these two boxes. The problem was they were very short of black boxy and hard to administer as well. Cesium is dangerous isn't, it catch fire if you put it in water? Good question so I went off to the time time an or axe at [unclear] lab in the US and it's because the the frequency of the atom is what is used to derive the SI second. The way it works they manipulate energy States, so already this is turning into a physics lesson. At the top is is a picture the cesium beam tube and basically lots of magnets and and this is used to basically train the oscillator that derives the one pulse per second time signal.

The source of the cesium supply is limited and after about eight years it's run out. And when it runs out it becomes inaccurate and it doesn't give out the right time so you have to keep these things uptodate and replace the parts. They are not particularly cheap. You also start becoming something of a Time Lord and I am not talking about wearing like bad scarfs or you know long coats or anything like that; but you actually end up starting to wear sandals and worry about that kind of stuff. So you actually start getting into all this stuff and finding out more about physics and atomic physics and time when we should have been con [unclear] on running fast gigabit Ethernet. This would give the wrong time if supposed to be a hyper arc rat clock if something bad happens becomes incredibly inaccurate so that was an announce with it.

We said we would replace it and went and had a look around online and found this company called Meinberg a German company, founded by two brothers. They are very well respected in terms of time keeping, definitely in Germany and mostly throughout Europe, these days. And they do this thing called secure hybrid system. It uses two time signals, one of this is signal from GPS using before and the other is D. C. F 77 signal, which is broadcast out of Meindfling in Germany. It's actually just near Frankfurt. It's the German standard time signal, administered by the the national physical laboratory of Germany, and there is US equivalents as well and you may be thinking why on earth are the Brits using German time. Well it's because it's German and it's therefore more accurate. And I am not...


MIKE HUGHES: And I am not kidding. It is actually more accurate. The signal uses this PZF modulation which basically uses this random phrase noise and it also encodes things into this so it allows you to get a high degree of accuracy compared to the a.m. only type stuff like MSF in the UK and people know it as rugby even though it's not transmitted from there any more. Convey leap second data as well and it basically gives you micro second accuracy, which is equivalent to what you get from the GPS time signal. When you configure the box depending on whether it's got an accurate time from GPS, to say if it's accurate from both you can be stratum one. If it's drifting beyond a certain amount we can then change how the box will respond and we can have it either stop serving time entirely and declare itself inaccurate, it can raise its stratum if necessary so that it, other times serves that are chiming against don't trust it as much, it actually becomes quite a powerful tool in terms of maintaining an accurate time service.

In terms of antennas, you both need to get them up on the roof and replace existing. Roof accessing colos is really good fun. Certainly there is a lot more stuff up there when we first put them up there. When we put this stuff in 99 there wasn't that much and the documentation wasn't that great because everyone could remember them in their head. Of course over the intervening seven years a lot more stuff up there now and which one is ours. It turns out that one of them and that is one of the others. It's simple sort of dipole type antenna and the GPS an a regular one and that is up on the roof on telly house in London. So we actually took pictures so that we have actually got documentation of where they are and what they look like for next time we have to do this. We also served time out of Telecity, for disaster recovery, if Telehouse went dark for whatever reason serve time out of another location. At the time when we were specifying this we couldn't get access to put multiple antennas on the roof, we checked to see if we could get a D. C. F 77 signal in the room and couldn't. This signal was not strong enough to get an accurate lock. As it was we could actually get a good MSF signal and [unclear] Meinberg had just launched this new MSF based clock which basically is a GPS and MSF. You can configure which one is the primary we are using primary time source with MSF backup and totally new product for them and they haven't done this before. And what that does is it basically just use one or the other, and there is an accurate oscillator inside the device and you can config how it falls over.

MSF is the sign for the UK time. That is not as accurate as the German time, allegedly, and that is actually referred to as rugby it, moves up to Cumbria just near the Scottish border back in 2007 when the contract had changed. That is just a very, very simple on /off key none of this PZF correlation, it's not got this micro second accuracy you get from D 77, the picture at the bottom is from one of the top of one of the masts that I found online. The box is really easy, they run Linux kernel, it's all NTPD in there, looks and feels like everything you are used to. You can even run the TCP and whatever else on there so they are a nice box. Not black boxy like the old things were. Very, very easy to set up as well, you get this sort of can you rememberers based thing for getting thing boot strapped and from there you use Linux to drive it.

It's accurate or no time. We can configure that. Stop serving it, measure things with SNMP things you can probe like what are the difference between the two time sources, obviously we hope that is zero, and from that I mean, what we learn more than anything. If you are running around cesium standards it can be expensive and you really need to be into this stuff. As it turned out we weren't really into this stuff and we ended up spending an inordinate amount of time trying to get into this stuff just so we could support it. So this was a really [unclear] option, the kit that we have got from Meinberg turned out [unclear] a really good option it doesn't require the sort of time an or access that you require to maintain something like a cesium standard. We will leave that to the guys in physical laboratories that live, eat and breathe atomic physics. Simple system is the way to go. Roof access and domination of the stuff on the roof in some buildings is is a real nightmare and the other thing is even though the kit itself was actually to use, we used the specialist local distributor that has got experience with mounting antennas and doing the cabling and all that kind of stuff that is relate today this, because even some of the stuff to do the cabling and getting the signal and the antennas in the right place, that itself is, again it's a learning curve with specialist skill. If you can get someone in to help you it makes life a lot easier and we had a local supplier to do that.

Anybody got any questions?

CHAIR: I have got one. How many wrong aerials did you cut.

MIKE HUGHES: We didn't, we found them fairly quickly. We covered an aerial up and went to see if the other lost lock and that did, that was one of our aerials. Covering another one up and nothing lost lock. That was someone else's aerial. Curtis?

KURTIS LINDQUIST: How much NetNod. We also run some NCP servers although a bit more complex than you do. We do have set of remediums and get signal from national time scales that we end up distributing which actually applications where they require you to be traceable to UDC, in router being traceable. Also draw back with using the we don't use, is that you can actually, you are at the hands of someone else of your accuracy, you no longer have control that have and there is also wellknown technique for jamming the signals or disturbing them to send into other times. [unclear] on reliability you might want to have control over this which is why we decided not to go down that road

MIKE HUGHES: We decided we didn't need to have control of the time signals, we didn't want to rely on SIG one, so I think this is where the SHS, the secure hybrid platform has good solution to that because the German government are in control of one signal or the British in control of MSF and we have got the GPS stuff as well which is fairly commonly used by a lot of other people so it actually does give us better level of protection than just trusting a single time source but without having to run our out reference.

AUDIENCE: Another thing, something we had been looking into do as well is interesting so see people find this interesting, is that we are provideing from these services locations higher accurate frequency sources for people to clock their transport networks, most people use GPS that can be jammed or buying it from someone else, normally at a fairly substantial cost and we can provide higher accuracy on this, get cheaper.

MIKE HUGHES: Yes. I don't disagree with that at all. That is a really good observation. From what we know people are using this for, it's basically you know synchronizing sys logs and just generally clocking their network to reasonable level of accuracy but not the level you were talking about. Any more questions? No. Thanks very much.


CHAIR: Next up is Kurtis.

KURTIS LINDQUIST: So, Kurtis from NetNod again still. This is going to be a short almost rerun of the presentation that you saw yesterday. Fergus asked me to say something about how we did numbering at the exchange points, if you see me giving this before go to sleep or read your email. Short presentation of how we did the numbering on exchange points for v6 addresses and what people have been looking at doing and why we did the way we did.

For us how to get exchange point addresses, you go to RIPE, they will hand them out, you [unclear] /64 for single or /48 for more LANS. We as I said before run several v. lance one for each MTU site and /46 for each of these v. lance. Out of the /44. We do IPv6 on exactly the same infrastructure as v4, it's the same investment lance so nothing special there. We decided not to do any special hardware or v. LANS for this. What are the option: Could theoretically use 64, it's a bit knowing just because you replace the hardware and you could assume you were using so many things do this, even so it's an operational nightmare to try to type all this long EU I64 numbers into your configuration files. Link local, same problems as BoF. If it works. And last is you have static addresses but what I mean is non-EU I64 addresses, addresses pre defined.

If do you this you take you /64 prefix allows to you do some end of end coding in the remaining bits, all these fun ones. Could do something more useful and coding people's AS numbers for example as the last bit. 32 bit AS inspection. We decided for this static addresses or nonEU I 64 addresses we simply gave every ISP a number, so here is is a colon colon 73, we used for that provider number 73 the older /24 and IPv4 and across all of the cities you operate in and they had the same number in all the /64s [unclear] v6 as well so easy to track who these people were and good example is put in AS number if you want [unclear] this just makes it easier to identify people.

That was all I had.

AUDIENCE: What are you going to do if there is no IPv4 any more?


AUDIENCE: I mean that your scheme, your allocation scheme.


AUDIENCE: Is based on IPv4 addresses.

KURTIS LINDQUIST: No. It's based on having a number at the end. That is not the v4 address, that is colon colon 73.

AUDIENCE: It's a loss of IPv4 address. Just number through.

KURTIS LINDQUIST: Just number through yes. I will have problems when I run out of the last 64 bits. And at that time I will be even....

AUDIENCE: Why didn't do you an action number scheme then.

KURTIS LINDQUIST: Because we had the numbers in v4, want today use the same numbers. Even if we start started from scratch we might have used action numbers, you can do some sort of more clever encoding.

KURTIS LINDQUIST: There was Lorenzo from Google what I am going to do if I run out of 264 members that is a good question. We actually happen to have the surrounding /23s, so if I want to could I actually number these a little higher. We haven't done that yet but we Co. I don't think that is an issue. We are currently after 50, more than ten years in operation so I guess that run out of v4 space before you hit that problem. Anyone else? OK.

CHAIR: Thank you, Kurtis.


GUARAB RAJ UPADHAYA: Thanks Fergus. Well actually, I am going to follow up on what [unclear] and Kurtis just talking about. After the last APNIC IX SIG there was quite a bit of discussion how do we allow IPv6 exchange points and there seems to be a lot of discussion going on and not well-established way to do it or well-established process to do it, both for ISPs, what do you need to put on the router interface connector to the exchange point. For v4 we have fairly well-established practice of blahblahblah, do this do this, v6 is looks like things are stilly involving and things are changing and so on and so forth. The idea is to produce BCP for both exchange point operators and ISPs, for connecting to an IXP so I am just trying to get feedback from the audience here, there is something the EIX working group would like to take forward. Just, this is just open question and discussion for folks here. Exactly the sort of discussion we were having like three minutes ago on why did you not use AS number in coding in the IPv6 address, and why use v4 address that is probably irrelevant in most cases, you can use anything you want.

AUDIENCE: If I think our scheme is more clever because it doesn't have any administration in it. People can figure out around IPv6 address.

CHAIR: That was Arian from AMS-IX for the record.

AUDIENCE: Our website has a fairly good config guide which is pub will I /KHRAOEU available and all the IPv6 options in there. It works quite well. Actually, I don't see there is any problem here.

GUARAB RAJ UPADHAYA: The AMS-IX guide is very well-known in the community

AUDIENCE: What are the problems with configuring IPv6? What are the problems you are referring to?

GUARAB RAJ UPADHAYA: No, the problem more than a problem, a lot of operators getting confused about what would be the best way to deal with this, we still had questions coming up.

AUDIENCE: I don't think there is one best way.

GUARAB RAJ UPADHAYA: Maybe this document, what are the current practices. This is open question as I said, we still, if it is feasible idea to document what operate verse already done and what are the common ways to do it. I mean it's up to the audience here.

AUDIENCE: I don't either think there is single best way of doing it but I do think writing down something on what the guidelines are, what are the different options you have and what are the benefit or drawbacks would be good thing to do for the exchange points. I am sure AMSIX can figure it out by themselves, people might want to think about it who would appreciate a paper on this.

AUDIENCE: I was a bit shocked we have this discussion right now in 2008. We already did this in 2001, I think:

CHAIR: We don't really seem to have a consensus on that, of going forward, anyway.

GUARAB RAJ UPADHAYA: Probably take this to the mailing list and send something over.

CHAIR: Yes. Thank you.


CHAIR: The moment we have all been waiting for. First I would like to say, thank you to Suzanne in an and her temporary replacement for scribing for the meeting, to our stenographer, and the Java scribe who enabled people externally to watch the meetings. So in no particular order...

(Results of the quiz) the main choice is fleecy or pair of cups and a hat and Tshirt for everyone Mary, the fleeces look nice a cap maybe too�

Any other business? See you all at RIPE 57 in Dubai. One of the comments was made, if you are going to go to Dubai try book early for numbers, flight deals are good just now but might be more expensive nearer the time. The meeting starts lo Sunday and finishes on Thursday, Sunday is a working day there as opposed to Friday. Thank you again.