These are unedited transcripts and may contain errors.

EIX Working Group session
16 May 2013 - 11 a.m..

CHAIR: Good morning everyone. Welcome back to part 2 of EIX at RIPE 66 in Dublin. The agenda has changed very slightly in that we did the first item on open IX at the end of the first session, so if you arrived specially to see that I am afraid you'll have to watch it on the web archive in a couple of weeks, however I think Marty and Dave are still here if you have questions.

The first up on the session is Bijal, give us the six month brief update.

BIJAL SHANGHANI: Hi, I am going to stand over here so people can actually see me. I am Bijal, euro IX. There is my contact details, also on Twitter myself and euro IX. Feel free to follow.

EURO?IX is an association for Internet Exchanges. Tempkins we have 64 affiliated IXPs, 52 of those are based in Europe operating over 100 Internet Exchanges. 22 IXPs are from the rest of the world. It was in 2005 when EURO?IX opened its door to not just European members but we allowed members from, well the rest of the world

Our newest members in 2013 are alibi IX, in Albania, TUNIX, KINIX in Congo, UAOIX in Dubai, Zambian IXP, CyrusOne in the USA and our first Canadian member, OTIX. We have 11 affiliated pay tonnes and those are over there. The newest once we have in 2013 are ALCATEL and HUARI.

Just a little bit more about the membership. As I mentioned the new members for 2013, some of these are are actually twins. And EURO?IX runs a twinning programme where the existing EURO?IX members can twin with IXPs in need, what they do is they provide support for IXPs in different parts of the world. And we have got two twins in there which is Zambian which is twinge with LINX, and KINIX in Congo which is twinning with NetNod.

Until ?? well this year, EURO?IX was ?? the EURO?IX region was sort of cut down the middle of Russia, but from 2013, we have actually extended the EURO?IX region, so now it's the same as the RIR region, which is ?? well, RIPE ?? RIPE NCC region, so now we are the same. And that means that there is I think Paul mentioned yesterday, 76 countries in the RIPE region and there is 76 countries in the EURO?IX region now.

So, news: My last update in Amsterdam last year I talked about the NIX federation, and in November last year in Internet Exchange we actually signed the MoU. LAC?IX and APIX, we signed an MoU and formed the Internet Exchange federation.

Last year in August, Af?IX was born, and it was ?? they are also going to be joining the federation and this is probably going to happen around, in the summer this year.

Other really big news is we're planning a collaboration with peering db, so this is actually going to be with the federation, and we're going to look at exchanging data between the federation and peering db, so, the idea being that the federation, the IXPs will maintain the data for ?? in the IXP database in the federation. And peering db will use the information from the IXP database and the IXP database will use peering db for information on networks and data centres.

And that's it. No graphs and no maps today. But, I'll just add that the traffic is going up and to the right anyway.

CHAIR: Thank you for that happy confirmation.

Any questions for Bijal? Thank you very much.


There is a note come through from Working Group Chair saying if you want to stand for the PC committee, nominations close at 3:00 tomorrow afternoon ?? sorry, this afternoon. And you'll find the link to it somewhere underneath the PC section apparently on the RIPE 66 website. So if you feel like helping put together the programme, this is your big opportunity.

Mark is going to give us an update on regional peering in the UK from LINX.

MARK STOKES: Hi there RIPE 66, as Fergus said my name is Mark; I work in the business development team at LINX. I am going to give you a quick overview of UK peering initiative. I have got some paper because my memory is crap.

I just want to go ?? it's no secret that we are setting up Internet Exchange in different places in the UK. I am here to explain why we're doing that. LINX's mission has been to keep traffic local. For many years that was from a UK perspective, and we have achieved that, what we wish to do now is to keep traffic local within the regions of the UK.

Obviously, looking at our EU neighbours, you can see that they have much more peering around their country, whereas the UK is focused on the capital, London.

All that peering in London creates extra demand for traffic and pushes LINX to chase the latest ethernet technology. So we're hoping that by spreading peering out in the UK, we can, you know, reduce that demand to always chase the latest hardware.

Anyone can set up an Internet Exchange. It's not hard. Making it successful and self sustaining is much more challenging. Using the existing membership structure means that there is no new legal entity, no new membership forms to sign. But, one thing we want to stress is that local ownership is key.

There is a bigger chance of success if it's driven by local members within that region. Another point is we're not going to connect these regional exchanges back to London.

Why not, you might ask. To connect the exchanges back to London would put us in competition with our members and it's just not part of our ethos to keep the traffic local.

So, we have completed a number of consultations in various parts of the UK, beginning in Manchester, at the start of 2012, ending in Scotland just at the start of this week. The consultations have allowed the local community to advise us and push the events forward. Without an active local community, we believe that the chances of a successful ISP are very slim.

IX Manchester is already up and running. Manchester has a history of peering, aka ManAp. Back at the beginning of 2012, LINX was thinking about regional peering, and it was actually approached by LINX members that are based in Manchester, and they wanted to peer ?? they wanted more options to peer locally in the Manchester area. So, LINX has the resources to help create a better environment to peer locally. So while we were preparing for the Olympics in 2012, we were also set something up a new IXP in Manchester. We popped the ?? the first POP we put a switch in was Telecity Williams, and since that's been installed, we have had Virgin Media have connected with a 10 gig port cable and wireless, and we have been talking to lots of existing members and we have had public commitments from BT, O2, talk talk to committing to connecting with 10 gig ports this year. We have also been speaking with Netflix and Akamai who are also keen to peer in Manchester.

The peak traffic in Manchester is reaching at 11.5 gig. And obviously the setup of a regional exchange is just the same as London. We have all the monitoring equipment, route servers, and everything that you'd expect from a well one IXP.

So, we have recently announce that had we are expanding IX Manchester, so we are going into Telecity's newest data centre in Manchester, Jewel House. We will also be going to the media city, where the BBC is, where Mark Kelly has a brand new data centre called Data Centred, and we are also going to be going to ICE co?lo which is run by M24/7. We have announced that the expansion is going to require a new hardware, so we are taking out the Brocade kit that we originally used and we are going to be taking equipment from extreme networks, so we can do cool things like VLAN tagging and doing connections, but as Bartek already said earlier, we are still waiting for the software on Extreme to catch up.

IX Scotland, this was the consultation, we had a second consultation this Monday, and this is ?? it was an easy decision. The business case and the technical benefits are very clear. Scotland is very keen to have its own independent IXP. So, the improvements in latency should be even more pro ups announced due to the extra distance from London and it looks like Polpul's south gile data centre is going to be the first POP. So, if you are interested in seeing the consultation that took place, the Scottish Government kindly recorded the entire event, so if you go on to YouTube and Google John Suiter, you can see the various presentations, or just send me an e?mail.

So, we have had a number of consultations in various parts of the UK. But we're not going to stick in an exchange where there is no demand from both content and access. As some people say you can't clap with one hand.

And that's it. Thank you very much. Are there any questions?

CHAIR: Thank you, Mark. Any questions?

AUDIENCE SPEAKER: Barry Rhodes from INEX. You say independent. How independent ISP independent? The LINX intend to walk away after a year, two years or so, or will it carry on?

MARK STOKES: We're in it for the long term. We are not going to judge the success after one year or two years, we're a non?profit organisation. It's not about, you know, it being self supporting within the first year or two. These things, peering takes time, you know, we have to look at this long term.

AUDIENCE SPEAKER: So it's not quite independent.

CHAIR: Well, I'll answer that, perhaps. My company is a LINX member, but we only peer at regional exchanges, and we benefits of LINX membership and local peering. You know, you can't turn your lines off without upsetting the members because auditor full member of LINX once you are there.

Wearing another hat. We'll probably peering in Scotland and that should be running in three months because the go?ahead was on Monday. And that will be Scotland's fourth Internet Exchange. Hopefully, this one will stay.

Any other questions for Mark?

AUDIENCE SPEAKER: How does that work with voting because you could be connected to IX Manchester only and you have got a vote on policy in LINX in London?

MARK STOKES: That's correct. Well what happens, it affects, obviously the Manchester members do get to steer what happens in Manchester. We are very keen to listen and for them to be the leaders in what happens in Manchester. They will get a vote at LINX events, just like any member.

AUDIENCE SPEAKER: Are you planning to separate that off at any point?


CHAIR: As a regional member, the value for us would be far less if it was a stand?alone exchange. It's the fact you are part of the wider UK community.

Robert was stirring it in the front row saying bring in independents. Post Scottish independence, there will still be organisations that are based in both countries, if it happens. It's not an independence issue.

I would disagree with what he said then.

MARK STOKES: On the Government side you could argue that obviously Scotland is not going to be independent in that sense, but as Fergus said, this would be the fourth exchange network operators like the structure, the mutual ownership of LINX, it adds that support, it's got the reputation that it's going to last long term.

CHAIR: You know, the Government is not ?? the Scottish Government is not involved in any of the running of the exchange or the setting up of, it they are merely facilitating meetings. And I think that's very clear both on the ISP side and on the LINX side.

AUDIENCE SPEAKER: Which will from non app. Aren't you supposed to be the Working Group Chair so area you answering questions?

CHAIR: Because when John got mark to do, it he asked me to support him on it because it's his first presentation ever. So... and I'm also reporting on the meeting at the LINX meeting. Well done Mark.

AUDIENCE SPEAKER: Thanks for a great presentation, Mark.

CHAIR: Thank you for that. Last questions, anyone?

Thank you very much.


Next up is Barry on hopefully something slightly less controversial, how to run an Internet Exchange using less software.

BARRY O'DONOVAN: Good morning folks. I am Barry O'Donovan from INEX. And I am here to talk about or properly introduce IXP manager to a lot of you. IXP manager is INEX as administrative and member portal. And since our foundation in the late nineties, we have always had some kind of admin crud interface as well as a member portal. But in the Nate noughties, we made a more concerted effort to build something a bit more professional, and a lot for futureful, and with the INEX committee's approval in June of 2010, we felt it was good enough that we would open source it as well to make it available to other IXPs. So that's now available open source under GPL to licence and there are a number of exchanges using it.

On the front end side it's a standard LIXP. And the libraries we're using are Xen for the controller, data network land two or the, Smarty for the templating and a God send for people like myself who couldn't design to save my life, boot strap from Twitter which makes it look pretty professional as well as J query.

This is is a screen grab of the admin interface looking at the AS112 customer or member. If you can't make it out on the overhead, just download the slide deck.

You can see we have a good overview of any given member. We see their aggregate traffic stats. If that graph looks a bit off, it's because it is. We have an ESXI service failure a couple of weeks ago we haven't brought AS 102 back up on the correct ports but we will soon that. Graph is a little off. The taps along the top here, we can very quickly at a glance manage a member's ports. Their users and contacts, CRM notes functionality and then some stuff RIPE NCC route server prefixes and peer?to?peer graphs.

From a member point of view, when they log in. They also get an overview tab, showing their aggregate graphs, some of our most recent members as well as any kind of news items we want to flash up to them. They can see their knock and peering contacts and update those directly in the port amount of they can manage ?? they can see their ports, graphs, they can dig into their graphs, daily, weekly, monthly yearly, as well as bits per second, packets, errors and discards. They can see all of their peering and LAN configure, IPv4 and IPv6 and so forth.

They have access to their own peer?to?peer graphs as well through the member portal.

Just some of the admin features. We have crud, queried update delete operations for managing members, user, interfaces, ports, VLANs, IP addresses and then stuff that doesn't change that often such as data centres, racks, switches and witch sports. We have full traffic graphs using MRTG as the pack end. That pulse switches and takes in packets and bits per second, errors and discards. We introduced peer?to?peer graphs using S flow late last year and this has proved to be a very popular feature. Where as we members logging in once or twice a month, we find members locking in a few times a week looking at their peer?to?peer graphs where they get to see their exchange of traffic.

There is some interesting stuff in the back end for us such as port utilisation alerts. If members start hitting 80% of their available capacity on any given port, we gets alerts so we can start chasing them for upgrades. We get traffic change alerts. If a member's traffic profile changes by a cert multiplier of their typical standard deviation, we get alerts, in case there is a problem or some other carrier problem, we can get them know that their traffic has changed quite significantly.

The most advantageous part of ISP manager is besides the front end admin into the database, is the all the scripts they back end that manage the auto provision the. When we add a member at INEX and new ports, at the become end the scripts come into effect and they start provisioning the BGP sessions and the route collectors, if the member has elected to use the route servers, this is pick this up, as well as reverse DNS or arpa entries for their exchange assigned IP addresses. All these has kicked in in the back end.

Some members co?locate equipment with us, but radius is update with their portal user name and password so they can access their consul ports. Similarly, all our members with a portal account have SSH access to our route collector so, that's a tack access updated for that.

From the member point of view, as well as the /TRAFB graphs and peer?to?peer graphs, each member can manage their mailing list subscriptions update knock peering contact details. The knock and peering details of the other members as well as what switches and ports the other members have. They have various different documentation such as pricing, procedures for connecting switches and stuff like that. And members ?? a member gets an admin account and they use that then to create additional users for within their own organisation.

So on the three points I skipped over there, I'll deal with those. The first one is peering matrix and peering manager. We snooped TCP 179 BGP traffic and we looked for conversations between members, so that we can as best as possible, assume that there is an established BGP session there. So we use that information as well as whether members used the route servers or not to build up a peering matrix for our primary and secondary peering LAN that. Information feeds into the peering manager and the peering manager will show you, on the slide here, the potential peers, it shows you the other members that you are not peering with on a given LAN using a given protocol, IPv4 or IPv6. That you are not getting on the route servers or bilaterally, so you can see in the first instance here, this is the AS112 member page, so for 3 Ireland, there is no v4 peering on LAN 1. But for say BT Ireland, they have a session on LAN 1 but they don't have a session on LAN 2, members can come in and pick up the members they are missing here F he hit the request peering button you get a POP up dialogue that will pre?populate with a standard peering request, that the include all your information, as well as providing information for the person you are requesting the peering with. So if it was a large organisation with lots of points of presence, they'll quickly see this is INEX Dublin, this is my peering IP address there, I can track that down. It's aware of whether you and the person you are requesting the peering is v4 and/or IPv6 capable. It will populate all the required information with just the required information.

There is notes also, you can put notice, if mum am people are managing peering requests, you can put notes and when you send a peering request within a system it will note that so if you come along to send it again, it will say look are you sure you sent one on this day, are you sure you want to send it again.

The second tab allows you, if you are so inclined, to pick up direct peerings with people even though you are using the route servers, so that will list people that we don't find BGP conversations between you and the other peers.

You then have a tab that listed all the people you are peered with and lastly there is a rejected ignored peers, so if you are trying to peer with someone who isn't going to peer with you, you can use a dropdown just to move that person's record over to the rejected ignored so it's not cluttering up your peering requests view.

The second tool that we launched just a couple of months ago was the route server prefix analysis tool. Now at INEX we use strict prefix filtering on our route servers using whatever IR D P database that you tell us to use. And from that we build up a prefix lists for the route servers for v4 and IPv6 based on your route objects, what I'm showing you here is a screen grab of a particular member who provides INEX with IP transit and on the search box on the right, I have just limited to INEXes AS number, so you can see on the middle tab, advertised and accepted, INEXes three prefixs are being advertised to the route server by this member and the route server is accepting them. In the first tab, advertised but not accepted, there is one prefix in there and what that means is that this member is advertising a prefix to the route server, but we're reject it go because we can't find a matching Route?6 object. The action here for the member is to go and add a route object to the database for that.

The third tab not advertised but acceptable, will show a member all the prefix that is they have registered but they are not advertising to the route server. There may be good reasons for this but it also may be a case that the member thought or felt they should be advertise ago prefix and they haven't and the action was there to go and update their prefix lists on the edge routers at INEX

This is a useful tool and it's especially for new members. It gives a quick way of seeing exactly what we are accepting and dropping.

There is a help screen as well which will show the member what AS number or set that we're use to go calculate this stuff so they can advise us if we should be using something else.

The big reveal is that INEX manager is coming to LONAP. At INEX we collaborate /REUT with LONAP and do a lot of different things. At a meeting late last year I gave an up /TKAUT on IXP manager and addey and Rob were in the audience and at the same time they were planning a schema to go and get a bespoke ISP manage for LONAP. We saw INEX manager and thought this will work for us. They came to the table for sponsorship and funding to get the extra bits developed, because now INEX are seeing a real turn on open source in this application. Some of the work we are just finishing for or LONAP is private VLAN support, which is something INEX needed as well. MPLS and private circuits. Additional CRM functionality. Big improvements in how we manage users and contacts. Contacts waned a lot for us, we weren't managing the contact lists. I go im/PW?FPLTS on that /W?R and we are allowing the members to manage their contacts lists and assign roles against the contacts. So that we have hopefully a much more up to date contact list for our members.

While ISP manager was not switch vendor specific, INEX used to be a Cisco house is now a Brocade house, similarly LONAP used to be Cisco and are now extreme. It was a bit of work to make IXP manager work for different switch vendors and how we added ports was /KWAOEURT flex. We are just about finished make that a lot more automatic and a lot more switch vendor agnostic. The next step is some migration scripts to migrate port data from the databases into IXP manager and hopefully we'll be ready for internal committeesing at LONAP by the end of May and be working with Rob to roll out to LONAP members during June and July.

If you want any more information, you will get Nick and myself at operations at INEX dot E I. That's the link where we have the IXP manager source code. There is also a mailing list if you are interested in subscribing. And thanks for listening.


CHAIR: Thank you, Barry. So, any questions?

AUDIENCE SPEAKER: Tomorrow Smith, wireless connect, RIPE /SREURPBLG inas well, just broke it there. Just a quick question. About the route servers, the auto provisions. Do you support any other vendors like mike tech or anything like that or have you any plans to do that?

BARRY O'DONOVAN: The route servers are Quagga and BIRD but that's route server implementation. Any BGP capable router will work with them. So...

AUDIENCE SPEAKER: The other thing you were saying you /OUS MRTG extensively for /TKPWRAFG and stuff like that. Do you support AS MP B 3 on that or...

BARRY O'DONOVAN: Our implementation is S MP D2.

AUDIENCE SPEAKER: At one point I started a small IX so I want to you know me so I can talk to you and I think I'll go ahead and implement this. Good stuff. Thank you.

CHAIR: Any more questions?

AUDIENCE SPEAKER: Emilio Madaio from the NCC. I have Mico Donovan from BT Ireland, he said I'd like to commend Barry and the team for IXP manager, invaluable resource. And ought IX is striving it and they are looking to extend to include multicast peering.

BARRY O'DONOVAN: The only thing that IXP manager doesn't auto provision is the switches, but we we do note whether a particular port is multicast enabled or not. I don't know if that answers his question or not.

AUDIENCE SPEAKER: Tom from IX Leeds. First, thank you. And second, we're looking at it too.

CHAIR: Any more? Thank you very much Barry.


Right, next up we have got an item on our IX inter?connection markets in Africa, Andrew and Michaela.

ANDREW OWENS: Good day everybody. My name is Andrew, I am from Teraco Data Environments in South Africa, and I run the NAPAfrica neutral peering exchange as well.

I'm going to talk about something that you guys might take for granted in mature markets, which is simple physical interconnect. And the effect that we have seen that that has had on peering. If I get lost, it's only because we only decided very recently to do this presentation. So, it's not the Guinness effect if I get lost.

So, you know, in how are history, an interconnect at non neutral facilities has usually been a sort of a layer 3 interconnect, so in a non?neutral environment, getting an interconnect between two independent companies means that you would usually have to go via the network of the host of the facility, so, previously before the neutral facility came along, that was how an interconnect was done.

So, we looking at this very specific example over here, where there is peering being done at an exchange point. So, you have a look at the bottom corner, we have got a provider C which is a very large content provider within South Africa, but they buy upstream from two different ISPs, technically they should be able to peer

What happened is there is an ADSL axis via one provider and they are hostings their content platform, which is extremely large at a different provider. There is a lot of inter?connection between their own customers so their own DSL customers would go to their own consent platform as well. But they were paying double for that. So they were paying around $70 per megabit for their own customers, their own DSL customers to get to their own hosting platform. The two providers that were providing upstream for them were exchanging that traffic at zero cost. So getting the benefit and, you know, charging downwards for that.

So per megabit, this large content provider was paying around 140 US dollars per megabit.

And this is an example that we have seen probably within the last two months. So it is still happening, and you know, most of this is transit, so it's layer 3.

What we have done with this customer at our facility now, is that because they have their own AS number, they can now peer directly with our open exchange. So, they still have the relationship with their two upstream providers obviously for breakout and for content that's not available at the exchange. But as you can see, you know, for a large portion of the traffic, they are now peering that at the exchange, and paying zero cost on the interconnect to the exchange.

So, from our facility, we don't charge interconnect price to say get to our exchange points. So that's all zero rated. And interconnects within our facility to different customers, we charge about 20 dollars for the layer 1. So the physical interconnect, because with neutral, we don't get involved with the actual traffic that goes over that, but your physical connection is $20.

So, obviously this customer is now peering on the exchange as well, which is something that they just didn't realise in the past that they could do. So it was a very interesting /STPHAR owe for us, just to you know, this curse /PHER is obviously now saving quite a bit on moving traffic between the two different platforms.

Okay. And we were warned against showing traffic graphs. So, this is not ?? this is up and to the right as per usual, but it's not a traffic graph. This shows the interconnects within our facilities over the last 12 months. It's interesting to note that we have seen a corresponding increase in the traffic at our exchange. Corresponding to the amount of interconnects that we have deployed.

This is ?? we can't name any names, but this is a financial institution that peers with us. And this is a graph that they have created, and what it means is they show the amount of money that they are saving on peering traffic. So, on settlement free peering traffic. So what you can see there is the two exchange points that they are at and the amount that they are saving per month, and it's quite funny, I think Andy Davidson came up with the new measurement which is called a gigarand, so we are not measuring in gigabit any more, we are now measuring in gigarand. So based on this by our calculation, this customer is saving about 25 gigarand per month.

Then just a little bit about our exchange. We consider ourselves the first neutral and open peering exchange in South Africa. Built along the same sort of principles that we learned from a lot of you. So, you know, within Africa, we make multilateral route servers available. We allow members to peer bilaterally as well and we have seen quite a strong growth in that.

Interesting to note, which is something that's always very interesting for you guys as well, is we don't enforce multilateral peering, but approximately 95% of our members have joined our route servers, which is very interesting. I think at the moment we have got a very small number that hasn't joined and even the once that initially didn't join, have been pressured to join by members that are already on the route servers.

So, it's I think a bit of a change to the rest of the world.

Currently, I'm not going to show awe traffic graph. But, we have just reached 70 members in South Africa. We are getting a lot of international guys coming in to peer with us as well. And our numbers are obviously a lot smaller than what you guys are used to, but yesterday for the first time we peeked at 3 gigabit, which is, you know, that's great for us, so it might seem a bit small for you guys, but we are very happy about that.

Just to conclude. You know, the key factors that we see to building an Ecosystem as we call it and what we have seen recently is neutral facilities are needed in our country. We see that as a key building block to better the Internet. We then ?? we also consider low interconnect costs are very important factor. We have had a lot of arguments with a lot of people about the cost of our interconnects. You know, as I said earlier, a lot of other facilities in our country charge very high rates, and typically will do it at a layer 3 level, or a layer 2 level, to interconnect different customers within the same facility. So, we see that as a key driver and we have seen that push growth of traffic within our facility is good cost interconnect. Obviously then, what we see next is the neutral peering, so, as I said, we have seen the correlation between interconnects and peering is very strong. But also, very important to note, the open peering, you know, concept. Whereas in the past, peering in South Africa has typically been a transit arrangement, so we have our incumbent who claims to peer with a lot of people, but it's a BGP session but they are selling transit over it, so it's not true peering and that message is something that we need to drive through the market.

And, you know, all of those things together we think are the key building blocks for an Ecosystem in Africa, giving a lot of extensive opportunities. That's it. Thank you very much.


Any questions for Andrew?

Mike is going to give us an update on the switching wish list.

SPEAKER: Hello everybody, my name is mike Hughes, you'll be pleased there are no graphs or slides. I am here, kind of like Olaf is forced to be here on Friday, in that I'm here on sort of behalf of Harold from Vienna and Martin from AMS?IX and myself who have been sort of custodians of the switching wish list and the idea of the switches wish list for those of you who you are unfamiliar is to help people that are running exchanges to go out to vendors when they are doing procurements and help in terms of evaluation and getting equipments got the right specifications. There was a review last year which, if you were here, you're aware of, and obviously what we'd like to do is do another refresher of what's in the switching wish list, remove any items that aren't particularly relevant any more, and also add any learning experiences that have come from the recent procurements that have been made. So, in the last couple of years obviously LINX has had a major procurement, DE?CIX have now had a major procurement and there is also some other exchanges have been changing their equipment. So if there is any learning experience that is people have gotten, it ?? please share them with us. What we'll do is I'll send the mail out to the mailing list with our contact details on, most people hopefully are the EIX Working Group mailing list so I'll send the mail out with our contact detail on. Our just grab one of us, we'll give you a card and we'll do that. What we'll do is we'll do a refresh of any items that people have got to bring. If anybody has got anything quickly they'd like to discuss now, we can do that. If anybody has got anything they want to raise now that they think should be put in it? And if not, we'll do it on the list. That's nice and quick. Thanks, that leaves us more time to hear about exchanges and people can do that kind of thing. Thanks.


CHAIR: I now everybody's favourite bit. IXPs updates. So, would all those who like to do an IXP update like to come up to the front. And we'll find out how many people we have got. If you come to the front, we'll come up one after the other.

First here, first on...

SPEAKER: I am now putting off the big hat from DCix ?? I am putting off the big hat from DCix, and putting on the smaller hat of UAE?IX. There should also be a presentation for that.

A few words.

ARNOLD NIPPER: UAE is up and running since October last year, the reason to build and exchange in Dubai is that there is a lot of fibres breaking out in this region, and also you have a lot of users here, about 300 million people. This was a region ?? the reason for us to establish an exchange over there.

The status is we have by now 20 customers, three more to come, this month content customers from the region, MBC, will be hooked up. Currently we have a switch in one location at Data Mena. There will be a second switch in a new location, Equinix Dubai, DX1, this this is be ready for service in the third quarter. Latest customer online peering was Akamai. And it sounds ambitious, but we are quite sure we will meet this target, our goal list is to have 40 customer to the end of this year.

And this what we're currently doing. That's it. Questions?

CHAIR: Thank you Arnold, any questions?

Next up is Harald from Vienna.

HARALD MICHL: Hello, since this programme item of the 30 seconds update I only reduced it to one slide. So, the main new things that we finally buying, we have bought a new platform. We have been very lucky with our current vendor and about just the biggest box they can offer. So this gives us the chance to make the upgrade until the end of June or so and then if with enough experience we try to start new services around autumn.

And something I would like to make you aware of is that the colleagues from v6 are organising an enter in Prague, so if you are interested in comparing the Irish beer with the very delicious taste of Czech Republic beer, you are invited to join us next Thursday in Prague. There is a web page link and I would be very happy to see some you there, thank you.

CHAIR: Any questions for Harald? Next up is Kurtis from NetNod.

KURTIS LINDQVIST: Hi, I am Kurtis, I work for NetNod. So, we are, those of you who are members of NetNod will have seen that we have spent some money. We are doing upgrade from the ?? to MLX 32 Es. This is so that we get a bit berth of utilisation of 100 gig port. We have announced there is going to be two service vendors for the replacement of the switches.

We are also going to upgrade the switch platforms in the other cities, got em bearing, and also the E platforms from the current RXs we use. This is also allows us to extend the reseller programme we have but not just in Stockholm but also the other cities which we occasionally get requests for. But that go be done after the summer holidays, or towards the end of it.

We have with any reseller, come core in Russia and we probably have a few more lined up to increase some of the presence from there.

And last, I actually had a question for members. We have been asked by some of the members that there is an interest as people are moving their voice interconnects over SIP, rather than to one this, over private PNIs, if this could be done on a dedicated VLAN and with some sort of QX markings. Would I like to see if there is any SIP on this and how great the interest in, I'll send a mail to the members list as well with you you can also talk to me afterwards at a social tonight or send me an e?mail if people are interested in this, we would like to hear about it. And that was it.

CHAIR: Thank you Kurtis.
Next up is Marty from TORIX this time.

MARTIN HANNIGAN: Hi, I'm Martin Hannigan from TORIX. And I have nothing different to say than anyone else that stood up here so far except that, so, this is it. Read the slides. I'm looking at the timer, I have got 30 seconds and if you have any questions catch he me after the break and let me get my picture. Thank you.


CHAIR: Thank you for another excellent graph free presentation.

Last up is JPNAP.

SPEAKER: Hello, I am as Fr. JPNAP in Japan, I'd like to give you about our JPNAP update this time. Total traffic leads to 400 gig this May. Mostly because of the growth of mobile traffic.

Second, a few member moved to 100 gig interface from 10 gig aggregation. Several users already use 100 gig at JPNAP. Last week we have replaced one of our allows servers from Quagga to bird to be more stable. That's all. Thank you for listening.

CHAIR: Thank you.

That's our last presentation for this meeting. Thank you all very much for coming. A big thanks to Amanda, Remi, Emilio and whoever else was doing our Jabber session earlier ?? Robert who was doing our Jabber session, for doing all our scribing. The minutes for the last meeting and this meeting will be going up on the website and we'll approve them next time. And unless anybody else has any other business, we'll say thank you and see you at the next meeting in Athens in October. Thank you.