These are unedited transcripts and may contain errors.
EIX Working Group session
16th May 2013
9 a.m.
CHAIR: Good morning ladies and gentlemen. Welcome to the first EIX session at this RIPE Meeting. A quick social experiment. How many people in the room are able to make both the whiskey BoF and the 9 a.m. EIX session? That's impressive.
We are going to have three presentations in the first session this morning followed by open discussion, this is a part of the meeting where discussion is particularly important and useful. So, please make sure you ask lots of questions to each of the presenters, you are very welcome.
Eileen is going to begin giving us an overview of the peering scene across Ireland. I'll hand over to Eileen for the rest of the welcome.
EILEEN GALLAGHER: Good morning, I am Eileen Gallagher from INEX and Andy and Fergus have very kindly asked us to give an overview of the peering scene here in Ireland.
We are going to cover off is the island of Ireland exchange that INEX has. Also one in Cork. And one in Belfast that there is a lot of energy going towards. First off we talk about C NIX, it was started in 2008 and it ran very similar so INEX and a lot of other exchanges in Europe in that it is neutral on the not for profit exchange. There is two drivers for the creation of it. One was that they wanted to peer traffic locally, they had some traffic there this they felt they could hand off very quickly, and easily, within the Munster region. And also that they wanted to create a purchasing syndicate for bandwidth and backhall to Dublin. The members do peer traffic locally. There is about 20% of the traffic that they have in total actually is peered within Cork, which is quite interesting. I spoke to the guy this there morning, Gerry Sweeney, and I was surprised it was to much. Backhall within Ireland has traditionally been very expensive. This has caused a problem for smaller members from the regions who really can't justify the back haul costs and then the inter?connection costs for INEX here in Dublin. That has stopped a few from joining in but some just get over that I suppose and they get up here. For some of them in Cork it was prohibitive. Their solution was a bandwidth buying syndicate so, six, seven of them got together, they decided how much bandwidth they needed and they actually went out to the market and they purchased it. That had a really good result for them because they were actually buying at much high are capacities and they had a lot more choice, some of the people who were willing to sell to them as individual members were not in the mix when this came to buying the bandwidth. They had got good pricing. They are peering about 20% of the traffic locally in the mustn't centre region. Knits Cork. What they do is that they aggregate the traffic under one AS within Cork at CNIX that, comes to Dublin via their backhaul link and they peer it at INEX, they have onward transit from Dublin. CNIX is the only INEX member, although there are six, seven members of the CNIX, the only INEX member that we have is CNIX so the individual members of their objection are not members of INEX so they have one vote. I think other organisations are looking at doing something similar to this, we figured that was the best objection option the at INEX we have just everybody is a member, in the same way has the same voting rights.
When the exchange was being discussed in Cork and they wanted some solution there, for the first year and a half or so all the discussions were around INEX will be come to Cork and put a solution here for us? And we discussed this and did many train journeys down to Cork looking at whether or not it was appropriate way for us to do T we were right resistant to T we didn't want to drop INEXes around the country, because that brings with it some problems in terms of local understanding I think perhaps is the biggest thing. We felt that the people within Cork, anyone who knows Cork, Cork people are ?? have their own ways about them. And I'm not being /TPAEU she issues there. And we felt that the local knowledge of the local wireless SIP industry down there and (ISP) local knowledge about how they wanted things to work for the future, the traffic and consent. They had a local TV station down there, and they wanted to peer that, here in Dublin we wouldn't understand how that worked and what the important times of days were for them or anything like that. So, our decision ?? one of the points there is that the backhaul to Dublin was one of the big parts of the mix. INEX didn't want to commit to buying backhaul, we didn't want to put that risk on the members. Members who perhaps wouldn't be peering in Cork. So we decided ?? and also that we didn't want to inter /KOBGT the two POPs. If we put a POP down in Cork and one it top, connects those two would put us on a train smash with our members and competeing with them and we don't do that. The decision therefore was made that we would support very heavily a local effort rather than deploying an INEX POP down there. We assisted them, I think very well, with developments, there was a lot of time given to it. We provided them with materials, like the MoU, the Articles of Association, our application forms, everything else for their use as they wished rather them having to go back and pay lawyers when we didn't have much money in the kitty. We welcomed their team at every INEX team and indeed all the members of CNIX are welcome. We want to create one community within Ireland rather than independent communities.
One of the interests is /PWAUG they are called the Cork neutral Internet exchange, when they applied to have their names with the registration offices they sent them letter saying no, they sent a letter saying you need to get approval from the other organisation that has a very similar name /TPOU which is INEX, we are so supportive them with wrote to the C R O saying we don't mind they can be called something similar to us.
We continue to support them F there is anything we can do for them, we will and that's how we see them progressing here.
Some contact details there should anyone have an interest in the Cork exchange. Really nice people, very happy to welcome anyone to Cork. And show them around their fair city.
Moving on now, Belfast hasn't an exchange at the moment. And I'll talk in main by INEX and it has a lot of members in the north and most of the eyeball networks up there we are hoping to bring on some of the U TV eyeballs as well in the next while under tie bus he is AS. Belfast is the last few months in the organisations up there have had and increasing interest in peering traffic locally within the area. They are not particularly set on it being Belfast and possibly somewhere else within the north of Ireland. There are some unique aspects to the north. There is no independent data centres up there yet. There is talk about them. There are a few smaller could he lows that are owned by various ISP type organisations. The next stage really for them, they had a meeting a while ago where they spoke with ?? we went up, LINX had set that up ?? about how they were going to approach this. They are setting up a community there now to formally look at how they are going to start peering within Belfast and going to progress from there. They have a lot of energy from Belfast City Council and they are quite keen really to get that moving. There is a lot of energy, I think they were impressed with how keen a lot of the members were up there for it to happen. As I mentioned the co?lo is perhaps one of the barriers they have but they are hoping to come across that.
Moving on to INEX T has five POPs in Dublin. The fifth one /WUR currently fitting out which is out at Vodafone in /KHROPB /SHAUBG, it's ran on Brocade turbo Ron for the 10 gig core. There is 3 dark fibre rings connecting the policy. That's ran on M R v? W D M equipment. As /PHEUPBGSed before it is an island of Ireland exchange. It's everybody from right up at the top of the country right down to the bottom. There is a lot of wireless ISPs in Ireland, and the vast majority of those have joined and those that hadn't joined are now joining because of NetFlix traffic. I get called about once a week about NetFlix, where it's forced them into a situation where they have to peer to get the best quality, because it is become something POP later
We are currently 76 members that, includes our associate members of which there are about 18. And we recently announce add reseller programme. That's come about, Ireland is an island, so we get to a stage where for people to actually become ?? to get new members we have toll abroad to get them. The Irish sea has has traditionally been expensive, less so now, and we hope that that downward pressure continues. We are hoping that the reseller programme is really going to encourage the exchange from the point that it's currently at where it's doing very well, up to the next stage that we start getting more eyeballs and more content from overseas and we have had great response from that. We have three, I think, three resellers that are keen and are currently actually selling services at the moment for us.
When we started in 1996, there were four members that came together to actually create INEX. In early 2005, there was 17 and things had started to stall. There wasn't really the money within INEX to facilitate growth, and the type of members needed a bit more work to actually get them on board, it wasn't just as easy as a couple of guys meeting, then they all decide to connect into the exchange, a bit more work and a bit more structure needed to be put around the chicks and that required some money.
We approached the IDA. The IDA had a report from Forfus, which is an organisation that looks at what the country needs for the future and where we are and what we need to get ahead. And for if I say had suggested that part of what was needed in Ireland was really world class Internet exchange. The idea therefore made a decision to give us loan funding, which we got for a maximum of 1.25 million. That was given on a very generous 18?year repayment plan and at the time a very good interest rate which is now just massive, because I think it was about 4 percent. At the time it was a great rate, but now it's quite ridiculous in Ireland because we are almost getting paid for money here these days.
What we did with that was invest in a second POP in a dark fibre ring. We upgraded the equipment and we also added members. So a CEO who was working very part time prior to that, Barry roads, actually came on in a more formal way, I joined the organisation and we also had an administrator because things like invoicing and stuff actually became slightly more needy.
Things went well. We actually made the final repayment on that loan last year in August 2012, it was fully repaid ten years ahead of schedule. Only 6 hundred thousand of the total loan facility was drawn down. So we had 1.25 million that we could draw?down but we didn't go mad. 600,000 was drawn down as I say, that's fully paid off and we're free of that. The expanding membership is now funding developments. Which is great, we're actually in a position where what we want for the few, we can actually get for ourselves. And it also means that we no longer have that cost every year of actually repaying the loan which means that we were able to put more downward pressure on the membership charge which is a very big part what have we do.
The relationship with the IDA also opened the door to several foreign direct investment companies. Because they were dealing with large companies, large and small companies coming in, mostly from the US I guess at that stage, although the focus has what I thinked to the Far East quite a bit now, they brought us into Google, Amazon, /KWRAO*. Microsoft came on board as the first large international company and gave us great support at that time and gave us, I suppose, a tick in the box to say yeah, this is all credible and it's going the right way and they continue to be a great supportive member for us.
So there is a large number of F D I companies are actually members of INEX, that continues, it's in a different way now because INEX is slightly bigger and it's a lot easier for to us manoeuvre through the market ourselves.
Maintaining focus is' important to us. We haven't sort of bled into other areas. We just do peering. We don't do any lobbying and we don't do any regulatory work. We help any organisations here in Ireland that dodo that, but it's just not for us. It's a large hole for money to go in if we started doing that in Ireland. We continue to build at world class infrastructure here. One of the things that we are doing is building up a very small reserve in order that we can fund the next stage of growth.
Innovation, I'll talk in a minute about the IXP peering manager that was developed. Developing the community for members and the IXP ?? the IP community in Ireland is important to us. INEX holds quarterly meetings here and one of the reasons that we do it is that many of the members that we have are from local Irish organisations don't get to go to the RIPE meetings. Don't tend to go to ICANN or NANOG or anything else that many in the UK would go to all of those meetings all around the world. A of the people in Ireland wouldn't go to those. So, many ?? you see many the eye refresh people here today you'll see the first time RIPE attendee badge on them. Many of them will be heading overseas for the next once. We hold the meetings to bring international speakers here and we are lucky we had a lot of support from people. If somebody wants to come and speak to the Irish IP community /W?R happy to talk to them about content and come over and join us. And we show people a good time. We don't have a fancy office. We invest the money in a great network and we don't have ?? there are very few of us. Most of us work part time at the organisation, but it works very well. We increase POPs rather than staff numbers. There are five of us in total. And there are no full?time employees at INEX; it just hasn't been necessary, and bloating isn't what we want.
What we do want is to give members IPv4 more less so we are increasing the members and routes. We are reducing the port and membership charges every single year, in the last I think five or six years, and I mentioned a small cash reserve earlier on to fund development but we have no intention of building up a large cash reserve. There is a graph alert here. I know that some you have are allergic to graphs. This one goes down and to the right. Like most IPX graphs F you look in particular at the 10 gig price, when we brought that in in 2009, it was 32,000 including the membership charge. That's now 15, so it is less than halved and our intention would be to bring that down further. It has to be cost effective for people to join INEX and to peer there. We are an organisation for the members, that's very important recollect and the members want lower costs and that's what we're giving them. And we stretch money very well. It's part of our cunning plan.
I mentioned the innovator earlier on. We have an open source provisions system and that was designed and developed entirely in?house mostly by Barry O'Donovan. It's both member and management faces. From our side we have visibility of what's going on with members. Whether or not they have accepted meetings, we have managed the meetings. For the members, it simplifies all that you are peering requests and management and it gives an overview of their peering relationships and matrix. There is also ebb?flow member?to?member traffic graphs available.
One of the things we have done is made that available to other approximations on an open?source basis. It's active already I believe Barry, with other approximations and it's also being considered and modified. That's something we want to push out and our cunning plan is to see that in as many exchanges that we can, because what comes back from that is good feedback and we can make it better for members again. We're very proud of that.
I mentioned about the various organisations that are members of the INEX, there is a short video here which I was a bit stunned by when I saw it... this really is an IDA video showing the various foreign direct investment companies that are actually here in Ireland. I saw this yesterday or last week for the first time. And there was a lot of these that I didn't know about. And it's very exciting. You look at some of them and think well they are there and that's great. But then I was speaking to one of the guys in Facebook the other day and I believe he said there was 400 employees within Facebook in Ireland. There are a lot of people here. A lot of people came in as technology companies, but also the med?tech area is massive for them. They are bringing in a lot of people. I think that's an impressive map and it certainly shows the cluster effect. What we're seeing now, maybe in the beginnings of it that when you get so many technology people in one city there is a lot more energy for people to meet on a more casual basis.
There is a movement between the organisations, so you actually get a movement of technology expertise and we are ?? there is one big reason why these people come in here, it's three numbers. 12 .5 percent Corporation Tax here in Ireland, which is a very nice number if you're a profit?making organisation. For comparison for other countries, those are the types of numbers you are talking about. The European average is 22.74 so Ireland is significantly lower than that. In addition to that the IDA provides a lot of grant assistance as well and a lot of assistance in you will actually helping the organisations start here, helping them find offices and everything else. It is a very well ran organisation I am very impressed with the way the teams work in Ireland and internationally. They give grant stuff for R&D, for capital investment programmes, training and employment, specially if you are basing outside of the Dublin area. One of the reasons that Ireland has proved popular is it's English speaking for US companies and there is a friendly corporate environment. We don't particularly take ourselves too seriously.
There are 200 ICT IDA supported companies here in Ireland, directly employing over 35,000 people which is a lot of people in a small country. And there are many, many more here that are not IDA clients. Recruiting talent is a problem in many countries, in the ICT sector. Addressing that recently it was announced there is a new tech visa, and they are giving priority to tech companies who are IDA?partnered companies who request a visa for somebody they want to employ. And essentially they are treating them as trusted partners, which is very good. It means that if they make a job offer to somebody outside of Ireland, that their tech visa is basically trusted so they are saying they do need them, they are the right person, we need them here, obviously notwithstanding whether or not they have committed crimes in their own country so normal rules apply in relation to that.
We are growing our own as well. I read very recently that in OECD countries we have the highest percentage of earning nearing students at 3rd level. So in the next four or five years we are going to have a lot of grad it's a coming out. Real support for technology companies, we even support with so many data centres here now, we support everybody with the weather. We keep the temperature really low here, it's something we have done in the last few years and you can see we are very effective at it. /*FP are
Fibre is good for us, /TPHAOEURD's not moving recollect that's a fact. We are right on the edge of Europe. We are the closest to North America that you can be when you are laying fibre, there is nobody can get any closer to us. Sea fibre networks in 2012 launched ?? there is a lot of fibre coming here or has come here recently. Sea fibre networks put in 72 pairs in Dublin to Holyhead last year. It means ?? that's quite a short distance to is means there is good value WDM kits that's making all that work and there is great connectivity from there on to the rest of Europe. Geoalso put in last year two diverse cable paths, one through Holyhead and one through D side. That follows the path of the UK?Ireland electricity inter?connection. So, you know, adding up to that that's three paths just last year. There are some interesting things going on here with power. We are actually in the process of building up a reputation of being able export power and power and temperature here for data all adds up to a very good future I suppose.
For the future: I think it was 2001 Hibernia Atlantic brought in the north and south coast routes from the US. They are hearing very good things about Emerald Networks and that seems to be progressing quite nicely. That will provide, we're told, the lowest latency routes from North America to Europe taking, we think, a few milliseconds off it, which is worth a lot of money certainly in the financial services industry. That will have 40 TB of capacity. Iceland Spur is included in that. A future branch to Portugal will actually provide us with possibly, depending on which comes first with that, and a sea fibre networks, the first direct route to continental Europe, everybody we have at the moment goes via UK. Arctic Fibre, this is planned to come through the northwest passage from Japan and the US west coast. That low latency route is showing a branch into Cork. Nobody can get any closer to the US than that. An interesting thing about that that that cable is coming over the northwest passage so, it's coming through safer waters from shipping, from sabotage, and from fishing. It's unlikely any divers are going to go down to get that cable when it's up in the northwest passage, or if they are I don't think they are going to succeed.
Hibernia, this is another one which ran into some issues in relation to the equipment choice they had. They have now actually given, I believe, the contractors certainly announced heads of agreement with TE Subcom. Interesting TE?subcom are are also the partner of the Emerald cable, so that could be quite interesting. We are hoping that comes soon. I put a question mark there beside the date.
Sea fibre networks are also talking about a Cork to France route. Which would be very nice. Hopefully that will come off. They have been successful with the route they did to the UK. They are in a funding phase at the moment. Apollo is also a potential there.
I think that's all on there. I know that there is a certain amount of scepticism when anyone talks about cables and we appreciate that completely. They are not real until they are in the ground and they are lit, or in see and lit in this case. There is a lot of talk though. There is at least five of these that look quite positive and I'm sure there is more there is a lot more fibre we expect into Ireland. There is great opportunity for Ireland for the companies that are here and also for INEX, you know, it works very well for us. With the low latency stuff, it means that possibly Ireland will start attracting more of the financial services trading platforms, which are at the moment there is a lot of financial services here but trading largely is pushed through London still. There is a lot of potential there.
We are very positive. And I'll end all of this with a George Bernard Shaw quote: "People who say it cannot be done should not interrupt those who are actually doing it." We're very positive. We wish a very happy rest of your time here in Dublin and thanks very much.
(Applause)
CHAIR: Eileen, thank you very much. I thought that was a great presentation. There is lots and lots of useful information there for somebody embarking upon an Ireland project. Thank you very much.
Are there any questions for Eileen in the room?
Okay. Whilst Bartek comes up I'll say one or two things. There is always a lot of people from the NCC who helped make the EIX a success, so thank you very much to Amanda who is making the minutes for this meeting and Robert who is relaying questions from the live webcast stream into the room. If people on the webcast have any comments during the presentations, you'd be very welcome to make that many through the Jabber system. Bartek, are you ready? Let's hear a little bit about what you have been up the LINX.
BARTEK RASZCYK: Morning everyone. I am a network engineer with the local Internet exchange. It's my first RIPE Meeting and the first EIX session as well so thank you very much for having me here. I apologise for my voice, it's not from yesterday's whiskey challenge, it's a genuine cold, so... I'll try to not cough.
I want to use this time to give you guy as short update about what's going on at LINX. It's a technical topics. As I believe my colleague mark is going to give another presentation later in the second part of this panel about local peering scene in UK.
First thing I want to give you an update on is 100 gig core update on the Juniper LAN so, we have a 100 gig on the Juniper LAN since before the owe limp pick games, so, we now want to introduce 100 gig in the core LAN as well. We are trying to move away from the 10 gig bundles because currently we are operating 64 byte bundle on our northern plain of the peering LAN and it's getting hard to manage, so, instead of the 10 gig links we are going to introduce 100 gig links.
Our core device that is we use in Telehouse north is Juniper PT X 5,000. We deployed 100 gig on the C FP based card which gives us I think, I believe it's two ports per pick so, makes four ports per slot. I guess we have got two peaks in one. This is allows to us dispense the optical transport key that we are currently using, we are using the M R v? driver kit and because we are putting a diverse dark fibre from Telehouse, it's in?house fibre, we don't really need the DWDM for this any more.
How does this look on the network diagram? We are updating the northern plain as I said. So those are red links that are highlighted here, each of them is 32 byte N so those are the links that will get upgraded to 100 gig. How does this look on the technical side. We are using the M R v? kit, so, going from left side, you'll have a Juniper PT X kit. Then on this segment here, we use grey optics, just a short range multi?node fibre. Then it goes to the MRV LD kit and then we put the colour optics and reverse on the other end. So this is currently two sets of 32 byte, which gives us 6400 gigabits between those two sites in total.
The second slide diagram, this is the same situation of the upgrade. So we can see we have removed the max.ing kit, we have moved the MRV kit and it's just dark fibre, 1 fibre per each 100 gig link and we are going to deploy 6 devices. This will basically increase the capacity in the northern plain of the land by twice, so we're going to have 1.2 terabit of bandwidth between the two sides.
Another thing I wanted to touch on is the extreme LAN. We have replaced the whole equipment for the Olympic games and now we just want to move that to VPLS. We want to extreme and did a really solid bog in February this year, we have tested the MPLS, VPLS, that they have on the X OS 15.3, which is the current latest release, and we think it's solid, it operates very well. We didn't find any major bugs. But for our deployment, for the LINX deployment, there are some specific things that we want to iron out.
So, for example, currently the hardware that we operate, the black diamond X8 and the Sam /ET 670, it only hashes on the MPLS label in the MPLS environment, when you are using lags, so when you have multiple LSPs, multiple VPLS instances, I guess that's fine but we at LINX we only have one VPLS instance and all the traffic, everything is in that one VPLS instance, so it doesn't give much entropy for the switch to hatch on, which basically resolves in a rather poor in a rather poor sharing on the end byte lags.
The way to fix that is to introduce LSP ECMP, so equal path cost?multipathing across multiple paths and we have tested an example of that code and with a 16 LSP going across the LAG the load balancing was working very well so we are just waiting for extreme to deploy that.
Another thing that we need to have in the X OS, that's the feature to bridge multiple VLANs into one single VPLS instance, which is currently impossible at the moment, and we need this for for connections programme to basically be able to resell the remote connections when we need to bridge multiple VLANs coming from our partner ports into the blings /P?PLS instance. Currently there is an limitation of one of this mapping, so obviously it doesn't work for us, but extreme, they have committed to deliver both of those functionalities in the next XOS release and that's a 15.4 and I think that 15.4.1 is due to release in late August, latest information from extreme says we are still on track, so we should be good and we should be able to convert the extreme LAN into a VPLS later this year as we had planned originally.
And that's it for my update so. If there are any questions...
CHAIR: Thank you very much. Any questions for Bartek? Thank you very much.
(Applause)
Next up is Thomas from the AMS?IX is going to talk about GRX and IPX.
THOMAS O'SULLIVAN: My name is Thomas O'Sullivan, I am a Senior Manager for Amsterdam Internet Exchange, otherwise known as AMS?IX, I am going to give a chat about GRX and IPX, this is not going to be an in?depth technical discussion but as we will see ?? this is not going to be an in?depth technical discussion but as we see as we go through along the slides here there is not going to be a who will lot of difference between AMS?IX's IPX and GRX setup compared to what happens in a normal ISP traffic.
So we'll just go back and we'll do a brief introduction to both of these. Let's start with GRX. This is a slightly less known AMS?IX service. So what is GRX? So, GRX actually stands for GP RS roaming exchange. This is the exchange of GP RS and 3G roaming data. So if you were travelling anybody, you 3G your home network, this is through this exchange it will happen. It's actually a private IP network, it's completely separate from the ISP in normal Internet traffic, and in fact it's limited only to the GSM operator community.
So, just as in the ISP community, the GSM A community also saw the need and the advantages of using an exchange to, a neutral exchange by AMS?IX to see if it would be cheaper and obviously for cost effective, so, since 2002, AMS?IX has been offering this GRX platform. Closed user group to inter /AEBGT all these mmm O, which is mole operator networks. Separately it's just a separate VLAN, it's separated from the ISP traffic. There is a GSM A supplied route DNS for this. Also, there is a requirement by the GSM A that all GRX members sign a memorandum of understanding, which is an MoU, which basically says that you will only use this GRX VLAN for GRX purposes and not for normal Internet traffic. And as of today, AMS?IX has 25 GRX providers, and if we look here schematically, this is how it looks without AMS?IX in the picture where the M N O Yours sincerely connect to the GRX providers and they connect the data between each other. With AMS?IX in the picture, it just changes slightly, where the GRX operators are actually interconnecting between ?? with AMS?IX for roaming data.
Looking at the providers, this is our 25 GRX providers we have right now. A lot of these names you are familiar with. A lot of telecoms companies here. And it's a good point to remember that there is 25 GRX providers, but behind those GRX providers, we have got the mobile network operators. So, these mobile ?? so in total they are probable hundreds of networks interconnected to these GRX providers which do the roaming exchange.
So, you may wonder how much traffic is passing through this network? Not a lot actually. It's pretty low. But it's just roaming data. As we look here we have got packet loss of almost 2 gigabits per second, but those packet loss occur during the summer months when people go on vacation, so people travelling, more roaming. Also at the end of year, at the Christmas time and new year, these are where we always see annually the packet loss of traffic. However however, if you put it it into perspective compared to AMS?IX normal traffic, an RSP provides a very, very small amount of traffic on the total amount of traffic on our knocks, I think like 0.1%, something like this.
But of course, globally, the ISP mobile Internet traffic is actually growing and we see here some interesting stats from Cisco, which showed us that there'd be a big explosion in mobile Internet traffic over the coming here is. I don't think that's any surprise to any of us. Actually at AMS?IX itself, the GRX traffic growth is actually twice that of our regular Internet traffic. So, we see that every year.
So, going back to the GRX providers, what does that mean for all of though guys and in the current climate with this growth, so, obviously they have quite a few challenges to face. The roaming rates are are dropping. And they have dropped a few years ago, you can see they are more ?? within Europe anyway ?? they have almost the same for all the Europe even country. I believe in the future they will continue to drop. The volume of voice minutes is dropping. I saw a survey recently from the US where the use of mobile phone for calling was actually fifth on the list of applications are now used by a mobile. I don't think any sympathise to anybody. I believe the Internet was number one where people used their mobiles for, on that survey anyway. A lot of End Users now use OTT players like Skype, etc., for calls, so, there is a drop in Revenue there. And of course the multimedia services. Like I say, there is a big explosion in that and how ?? on the from the GRX providers' point of view, there is a desire to have quality of services for these time critical applications that they provide. And of course, now, with the future of the roaming, the rollout of LTE, there is a desire also for quality of service and these prime time critical applications.
So what happened was that the GSMA community and the association have evolved GRX into IPX. So in 2007, the GRX defined the IP exchange, which is now called the IPX. They wrote, of course, a lot of documents and guidelines for this, as you may be aware. At least I'm aware of four of them. The one that was most important and that was most relevant to AMS?IX was the IR 34, which is the backbones guidelines. What are the changes between GRX and IPX in these cases?
So GRX introduces additional stakeholders. There is people like content providers, application service providers, fixed network operators. They cannot ?? important point here /?BGS they cannot directly peer on an IPX exchange, they are always connect via an IPX provider. That's current status anyway.
The IPX exchange, or the IPX itself requires end?to?end SLAs and end?to?end quality of services. This is for cascading SLAs and billing. Within the IPX is actually already differentiate different classes of traffic as well as different service communities, so, what does all this mean for AMS?IX?
Well the same at AMS?IX again, we were approached by a lot of these IPX providers asking if we could facilitate an IPX peering platform on our network. We wasn't ahead, did some trials, we looked at the documentation, we saw there was no additional implementation or technical requirements from our side. We did of course have to install some probes, etc., to monitor and show that we were within those SLAs. So, from an AMS?IX perspective, the diagram looks thick this, which is actually very similar to the GRX diagram I showed earlier on, and that's because from an AMS?IX perspective, we didn't do a lot more. We didn't do any change technically on our platform for these people. So, we implemented an IPX peering platform. We have a separate VLAN, we have a requirement for two redundant, separate connections, ports per connection on separate colocations. We introduce an SLA on our platform from within our platform between other edge routers and we set up probes on those access routers for the measurement of the SLAs.
So going back to the IR34, we actually had to look at a document and see what was in it and what specifications they were talking about. So, one of the things that is an IR34 and I don't want to go into depth into these documents because it's not part of this forum really, but just to give an explanation of how AMS?IX took, looked at the document, how they implemented the applications.
So, here we see that there is a load of different service classes. And for the IR34, and they have given different quality of services, depending on what type of traffic class that we're talking about, so they have given like for instance, video and VoIP is the highest because that's the most time critical and where delays and jitter are most important, and they have rated them all down. So from the AMS?IX perspective, we are not involved within what happens in the IPX, so we just took whatever the highest service class was to look at what quality of services they were offering for that, so we took the conversational, and within the conversational, quality of service parameters, they have got service availability, packet loss, delay and jitter. And from this ?? and these are the definitions for these specifications. So the availability is 495s, the packet loss is 0.1% and you have got jitter for inter or intracontinental. So, within the AMS?IX platform, we implemented the IPX exchange like this where you can see here on the left?hand side, you got the one IPX provider, they have the two redundant links. On the access routers, we have the probes set up so we can measure within our own network via to the other IPX provider that they are going to connect with.
On the overall IPX ecosystem, the scenario likes like this this where now we have service providers which could be actually mobile network operators. They will connect via IPX providers. You still have our GRX only providers involved. And in between, we have different service communities, these might be signalling such things like Sigtron, which is SS 70 IP for the latest maybe diameter for LTE. Also, IPX transport or IPX voice, so this is how the whole IPX ecosystem will look like.
So, we launched this in 2011, a lot of fanfare, publicity and all this. So what happened since then?
So, actually not quite a lot. We have eight voice providers connected to our platform. Quite a slow process. We are dealing with incumbent industry where old business models are still prevailing. It takes time to make decisions. And we are hoping, of course, that the LTE roaming rollout will actually put some momentum into this. We are actually expecting some more of course soon.
So, but actually also, since this week, we have also launched ourselves the inter IPX in Amsterdam Hong Kong and AMS?IX Hong Kong and AMS?IX Caribbean. So we are very positive that this will take momentum and logical grow according to the LTE rollout.
Thank you. Any questions?
CHAIR: Really interesting stuff, thank you Thomas, any questions in the room? Nina has a question.
AUDIENCE SPEAKER: I am Nina from TDC. It's not really a question. It's more ?? well, it is a question. So, when I look at the ecosystem slide that you had there, what struck in a my mind was hey, apart from the Q S thing that's the Internet ten years ago. So, do you have any idea if this will actually take off? Because, what this assumes is that people like, well my friend over here, Dave, would connect into IPX to reach the mobile End Users at the service provider, but that does not scale at all, because with the traffic, why would he pull ?? why would my mobile company pull the traffic from the exchange in Amsterdam where I might have have a direct connection into the mobile back endnd get the content from there? It's way cheaper, and it works as well if I have the server standing next to my back end any ways, so, do you really think that this business will take off?
THOMAS O'SULLIVAN: Well, that's a very good question actually. It is, of course, a long time in the making. This is a very old diagram, as you actually point out.
AUDIENCE SPEAKER: I tried to figure out what that was in 2008?
THOMAS O'SULLIVAN: Like I was saying, that this is the industry itself of course very, very slow. Sometimes they are going two steps forward to go three steps back, and that's what we're finding. We find that decisions are taking ?? takes a very long time, meetings go on on forever, nothing actually seems to happen. So, while we do have some IPX providers, it is a very slow process and there is still a lot of question marks on it, will it actually fully rollout? But, as I was mentioning earlier, there is still a lot of pressure of course within the community saying well with all these OTT players, etc., should we do something within the industry? And that's from the telecoms themselves, to actually catch up basically with the rest of the Internet community?
AUDIENCE SPEAKER: But what these guys at the telecom themselves are missing is that they have the really cheap transport for that content right next to the back end in their own network. It's called the Internet.
THOMAS O'SULLIVAN: Yeah.
AUDIENCE SPEAKER: So... I just ?? I am still confused that this is still trying to work out.
THOMAS O'SULLIVAN: What I wanted to point out there is actually from AMS?IX's perspective, we actually have done nothing extra in our own network for these guys. We actually treat their traffic the GRX and the IPX exactly the same as we treat normal ISP traffic. We don't do no marking, we don't prioritise the traffic. So that's why we say well ?? you can say what the difference S but of course from the GSM A perspective, they say well, if we are dealing with voice, we are dealing with time critical factors, with delays and jitter and and security, and so, from their perspective, they say well, the Internet doesn't actually bring us that, so we will have our own alternative to that, to the Internet.
AUDIENCE SPEAKER: That actually reminded me of the question that I actually had, which was how do you implement your calls service over the switch? That just /OELT fashioned over provisions or do you actually have traffic classes because I noticed you said ?? he look at the best class, we look at best effort but some of the other couldn't /TPEPBT might actually not work well in best effort.
THOMAS O'SULLIVAN: Like GRX itself, GRX is best effort. Whereas the ??
AUDIENCE SPEAKER: Not best effort. Expedited ??
THOMAS O'SULLIVAN: The reason I mentioned about the traffic classes, which is a good point is like I said, we didn't do anything extra, so what we did was we looked at the GSMA recommendations and we took the highest service class, the traffic classes that they were specifying, took their highest specifications and just put probes on our network and said, okay, we measure from this axis router to this axis router, are we within the specification? And yes we were and we did nothing extra, we just put two probes in.
AUDIENCE SPEAKER: Excellent, that answers my question. Thank you for doing it that way.
CHAIR: Great questions, any more questions for Thomas? Something I didn't quite understand. In the previous model with GRX, you had to be a member of the GSMA club to participate. The new standard IPX appears to be more content to mobile company delivered. Has the participation in IPX therefore been opened up to non?telco content providers or is it still a club?
THOMAS O'SULLIVAN: That's actually a very good question as well. So, within the IPX community, of course GRX is now pun of the service communities. So, the idea would be that from the GMSA perspective ?? if you were not using GRX, you should not have to sign a document, an MoU regarding GRX. But obviously if you are using GRX as part of your IPX package, then you would have to sign the document, because even within the IPX ecosystem, the GRX is actually still separate because it's actually a lower class. So... yeah... but all these things are still, I can say, still up for discussion in the GSMA community.
CHAIR: Very interesting and I hope that while we are discussing these things a lot of them discover the Internet as well because a lot of these problems are solved out there. But it's interesting to see what another community does for inter?connection. So very much appreciate you presenting this. Thank you.
(Applause)
Okay. We have about half an hour more before the break. The next presentation is from Greg Hankins from Brocade who is going to tell us more ?? whilst a lot of the Internet exchanges in the room, Greg is going to shock us all with a story about 400 gigs.
GREG HANKINS: I want to start off with a little bit of history and an interesting I had it bid an ethernet is actually turning 40 next week F you look at the quote in the red box, this is something that Bob Metcalf wrote on May 22nd 1973, this is kind of considered the starting, or the birth of ethernet. So, 40 years later, here we have and then if you look at this diagram, the network that they sketched out had ethernet kind of everywhere and it turns out that's actually what happened and we do see ethernet at several different points. Some of the things that didn't quite work out we didn't see the telephone eater really develop, but certainly for wireless and wired ethernet has become the technology of choice I think for network operators.
On to 400 gig ethernet. I'm going to anti?obvious question first. Why 400 gig? Why not /TER a bit. Well terabit ethernet currently with the technology that we have today is just impractical. Begin that we couldn't do terabit ethernet we had two choices. The first was to wait and not to do anything. And the second was start working now on something that we actually can do which could be ready in a couple of years when the market needs something faster. So there are a couple different IEEE study groups that did bandwidth assessment needs from network operators and analysed all the options so that when we started the next bit of ethernet at 400 gig, we went in knowing the speed in advance. This was different than in 2006, 2007 when we started working on the higher speed ethernet which was still up in the air at that point. So we went into the IEEE with just the objective to develop a higher speed ethernet and that ended up being two speeds. With this next speed we are going into the IEEE with the objective of developing 400 gig from the very beginning. This means we'll have a standard much quicker.
If we look at what a terabit ethernet module could actually look like. In the engineering words from some my colleagues, the module would be gigantic. So that's a technical term. I asked him, okay, can you give me more detail. It turns out that it's based on how much power we can cool. So we can cool about 1 watt per square inch of form factor on the module. If you look at just some extrapolations. If we have a CFP2 module that takes 12 watts and we have to cool about 120 watts, then it would be about six by ten inches. So that's a gigantic module. It takes up over 25% of the board space and impractical to develop something like that. It would use ?? if we used 25 gig signalling, it would use 40 channels and that's impractical to put on the board in terms of signalling.
So, on to what we can do today. The IEEE CFI, this is what kicks off the project in the IEEE, was introduced last March the study group was approved and they are meeting this week, so we'll see what they come up with. They are going to define the reaches and media. So as they said the speed is already decided to be 400 gig and there is also an MSA, which is a multi?source agreement. This is for optics where a bunch of people get together and say they are going to build this optics to this specification. That's going to be called the CDFP, and we expect the first interfaces to be available time in 2016. Much quicker than the 1400 gig standardisation process took.
We have the same thing that we saw for the previous speeds, so for 10 gig and 100 gig, we expect a couple different generations of modules. The first generation is interesting because there are some proposals that you see in the middle to actually use 400 gig CFPs to do the cabling and the signalling, so this would somehow be bit striped over 400 gig CFPs into one multiplex high?density cable. This is interesting that we see as we run out of board real estate and just front panel density to do things with optics. We are coming up with new high density cabling options that could be interesting, I don't know if this works in practice but anyway it's a proposal.
The third generation that you see could actually use four times 100 gig electrical signalling, that's expected to be around 2020 sometime and this is when terabit ethernet actually gets feasible. So you can see that each speed of ethernet actually builds on the previous generation, so 100 gig used a lot of 10 gig components. 400 used a lot of 100 gig components and terabit would also use a lot of 100 gig components.
The CDFP is interesting. There are a couple of prototypes that are being kicked around in the MSA. This again relies on high density cabling, so this is multi?fibre ribbon cable, so 16 transmit and 16 receives, we are moving away from, because we are just unable to multi?flex all of the waves on the one fibre. This actually uses 16 waves of 15 gigs. It's impossible to build a multiplexer into the optic that's cost effective and actually a reasonable size. We are moving to parallel tables because we just can't get the capacity and the density on the panel.
Then this is what a switch could look like in the future. Right now we have the cast of about 400 big per slot or maybe 800 per slot. This new CDFP would allow to us stack two optics on to much of each other and possibly use high density cabling or connector to actually get 64 fibres so, could you get 8 hundred gigs into an interface on a one use switch. So that's pretty impressive high density that we are going to be able to do in the next couple of years.
There is a lot of information that you can read up on, so these are all the IEEE study groups that they have and then the last one at the bottom is the call for interest presentation, so this is the presentation that kicked it all off. You can see some of the background and some of the data that they brought in which showed the need for higher speeds and things like that over the next couple of years.
That's a quick update.
CHAIR: Thank you very much Greg, any questions? Everyone is still very tired today. We have a question.
AUDIENCE SPEAKER: Sebastian. Just one quick question regarding the multiplexing, so we are using now 32 fibres or 64 fibres/, is there any research going into putting it back on one fiber? Is there something that is even on the horizon? Because all the long distance cabling will be very much ineffective with this kind of multiplexing.
GREG HANKINS: That's a great point and you are light that over long distances this becomes a challenge. This specific interface would be designed for shorter reaches so something up to, you know, maybe 500 meet the would run within a data centre. A short reach single mode application. The problem that I touched on briefly is that if you imagine that these are 100 gig CFP and not the 400 gig CDFPs, when we enter grated multiplexing piece on the optic, it made is very large, because we had to integrate those components. They are getting smaller, but you know that's the challenges that we have to have the actual, the WDM component inside the optic to dot multiplexing so it makes it large and complex and more expensive.
AUDIENCE SPEAKER: Thank you.
CHAIR: Any more questions for Greg? Question, I see Blake going to the mike. Hopefully you are going to the mike.
AUDIENCE SPEAKER: Blake Willis with Neo Telecoms.
Do you see any movement with some way towards to be able to add anything to the MSA with regards to the digital diagnostics after built in passive filter in MUX D MUX inside the optic, we had a lot of issues with 100 gig where there is been an issue in the, you dot fabrication or something with the passive component and you are just completely blind to that because you can't field telecommunications just have a power metre, put it on there. Okay, it looks all right because two of my four channels work and the other two don't, etc., etc.?
GREG HANKINS: That's an interesting question, and an interesting problem. I don't know if anyone has ever thought of that, but I can send that back to the people that are working on the MSA and ask them to think about that.
AUDIENCE SPEAKER: Francesco Chiappini, Telecom Italia. I have a practical question of still regarding the connector. How do you ?? do you think we will have ever see patch panels and the ODFs for that kind of form factor for the fibres?
GREG HANKINS: No, I think you are talking about this one. This would eventually go into something called an MPO, or an MTP, which is already in use today, it's a high density, they have them in 12 or 24 or 36 pair fibres, so we wouldn't see patch panels like this but what have today is horizontal cable that goes into an density connector and is broken out into whatever ??
AUDIENCE SPEAKER: So, there will be sort of an interface and then you have a bundle of 16 or 24 fibres, whatever, and then traditional patch panels for ??
GREG HANKINS: Exactly. And this is how what 100 gig is using today, so the 100 gig short reach interfaces and actually the 40 gig short reach interfaces already use MPO cable as the cabling that goes into the optic. So...
AUDIENCE SPEAKER: Mike Hughes, this is more of an observation about Blake as comment or question rather than anything specific to Greg. When you get 100 gig inside your network, if you haven't had done an UDM on your network before it's crept up on you by stealth. Sending field tech out with power metres to die knows this stuff sin sufficient and you have to send them out with OS As basically because that's what's happened. You are blind because you have not got an out board transport element, everything is in the 100 gig optic and hidden from you. So you have got to go and bring some out board test gear to basically deal with that. But that's certainly my experience anyway. That this stuff creeps up on you and you need to actually have the right test gear and you have got to budget for that as part of a deployment as well. It's not just a case of buying a line cards and putting them out there you have got to budget for having the proper test gear, otherwise you come unstuck.
AUDIENCE SPEAKER: The particular issue we saw was we have a lot of experience with WDM, but with traditional WDM, you can unplug the patch cable from the optic and measure it, and with the CFP, you can't.
AUDIENCE SPEAKER: You have got to look ?? to be able to use the OS A and look out at the MUXed out all at once, you couldn't look at a single waive of the output.
GREG HANKINS: Hopefully, if we could improve /TKOPL that would help some too.
AUDIENCE SPEAKER: Absolutely. I'm not saying that's an invalid comment. That's a perfectly valid comment but if anybody in here is thinking of doing things with 100 gig this they weren't at the moment, it's something that you need to go into with open eyes.
CHAIR: Any more questions for Greg? Thank you very much. That was a useful discussion, thank you.
(Applause)
Next up we are going to have an update on the open EIX project, presented by Marty Hannigan, this was originally scheduled for after the break, but we are bringing this forward to allow for more discussion time after the break. So Marty and Dave Tempkins please.
MARTIN HANNIGAN: Hi, I'm Marty Hannigan. What is it? It's a standardisation of Internet expense. Just to give you an idea of the actual problems. When you look at data centres and occupying space with your equipment, they are pretty much key components in the real he is State cost, he be gee and interconnect cost, in the States in particular, if you kind of set some boundaries and you look at something like a 2 meg watt facility, you are looking at some average number, so, let's kind of use round numbers and look at the typical expense, 100 interconnects kind of across a yearly recurring revenue would cost around 600,000. When you up that to 500, you are looking at 3 million and when you up that to 1,000 you are looking at roughly 6 million. When you aggregate those numbers out over pretty much the US metro, the 15 markets and you carry that out over ten years, you are looking at nearly a billion dollars in revenue or expense and that's just at today's cost of capital the that's the problem that we're working on or we're trying to solve.
So, open EIX is a community drive to effort. It was organically developed. Dave was the creator, he sought some volunteers, we jump you had on board, it's open to anyone, any EIX, any data centre operator. It's not a contract vehicle, we are not going to be cutting deals with IXPs or people that want to appropriately an IPX, we are simply developing the BIS case and acting like match makers for the most part but protecting the community's interest and it's focused in North America for you now.
We have built a group of folks to do the work, consists of CDMs, ISPs, eyeball networks, data centre operators.
We built a community governance model, how are we are going to operate this thing. We have support from the community which we are going to demonstrate in a minute. And we think we have created an competitive model for interconnect which we think will help solve some problems.
The OIX group. John from Google, Eric from core site. Myself. Tony from telex, Dave from NetFlix. And BArry from Telecast.
If any of these folks are in the room could you raise your hand. Just so...
I guess it's just us today. It's a wide variance of people. We did surveys, we have responses. Dave did presentations at two NANOGs and a few other venues and there was some support and some support to continue to working on this, which we did. And to under score that support we intend to elect members to OIX to solidify.
So, first kind of survey data that we can present is what the actual need is and we surveyed the community, mostly North America through the NANOG list and that is what we came up with, northern Virginia obviously is the kind of number one indicator of interest. Followed by New York and Silicon Valley. We kind of new this but this underscores that as well.
Here is the criteria. We did a survey, we call agree, preferred markets for support. Minimum service offering, established an IPX governance threshold, basically with what we don't want to do is create something that will grow out of control and just take us back to the point that we are already in. And so, what that entails is an endorsement kind of proposal where in exchange of agreement to operate and open IX endorse facilities, I think ?? we did a survey, we established what markets people want to see the project operate in. We set up a minimum service offering, we established a requirement for a governance threshold, and in exchange for an endorsement, you'll agree to operate an open IX endorsed data centre. We also here is the survey of responses for the minimum service offering.
So clearly the support for a standard Internet exchange for VLAN product doesn't look like there is a whole lot of interest in wavelengths but it's there.
The governance model. So, we want to avoid creating, you know, a monster for the most part. So, we have cod /TPAOEUD the engagement requirements. There are some stakeholder accountability. There is collaboration of course with the community. And there is a requirement that in exchange for the endorsement that you'll be organised in the United States as an IRS recognise niced non?profit.
How should the IX be structured? Non?profit. Obviously, clearly the winner. And pretty much in line with what everyone expected. So, the data centre side of the equation. We did do a survey but we know that we need data centres to provided a /KWET rack space for an OIX, we need them to provided a /KWET redundant power. We need low lost no cost interconnect on the metro footprint and they receive the endorsement and the matches are made in exchange for that.
How it works /KPOR IXP, you submit a proposal to the endorsement committee, there would be a review, a feedback loop. There would be a public comment period, there would be a post somewhere that would say, there is a proposal for an IX in one of the preferred markets. Basically it would be approved and endorsed if it met the criteria and the app wasn't withdrawn. Theoretically the comment period would define what was need was in the market and if there were any kind of problems with the application. I think that a proposal would likely be endorsed whether the community thought it was a good idea or not as long as it met the criteria, but obviously with the public comment period, if there was a hugely negative response, it would be fairly difficult for someone to move forward even with an endorsement and it would be fairly difficult for them to get an agreement with the data centre if they, if there was a pre?failing negative aspect to it.
Same thing for the data centre. It meets the minimum criteria. Proposal submitted, again a feedback loop, another comment period and then there is an endorsement.
The final bits. What's left? We have some draft documentation to clean up and take out of beta. We need to figure out how to provide the comments period. And implementation. So, we have done quite a bit of leg work on this. We already have proposals that are willing to be kind of submitted to the process, and we already have data centre operators that want to be endorsed as well. We think there is a fairly good amount of traction here and we are interested in additional comments. We want to know if you think this is in line with expectations, if there are some suggestions, if there are some flamage to be had. And over to you.
CHAIR: Any questions?
AUDIENCE SPEAKER: Tom Burke, so cable charges and data centres just basically money for nothing, why would they let you set that price or undercut it?
Cable charges and data centres, that was what you were talking about earlier, right?
DAVID TEMKIN: So to some extent. We are talking about the cost of interconnecting to the IS fabric, we are not looking to build cross?connect bypass for member?to?member organisations.
AUDIENCE SPEAKER: This is /AR from the Internet exchange. You mentioned the data centre has to meet the minimum requirements. Are this list of the requirements public or could we have a look at it because it would be nice to see what your minimum requirements are and I can look if my data centre also would comply to this list.
DAVID TEMKIN: It's pretty straight forward. The idea here again is we could have started in multiple ways in this project. A lot of people ways that people start in IX platform they go out and they buy a switch or they get a switch donated and they install and they hope they can build the policy around it and get people interested to make it work. Generally that doesn't work. There is a lot of lonely switches sitting around the world doing nothing because of that. The idea here was to provide a reboot so that we create the framework, the requirements are relatively sparse to allow for the fact that different people have different business models. But to provide a consistent product across data centres we need to know that we have got these base can things.
AUDIENCE SPEAKER: Thank you.
AUDIENCE SPEAKER: Which will. Just to be clear are you guys providing a switch and running a switch?
SPEAKER: No.
CHAIR: Any more questions for...
AUDIENCE SPEAKER: Mike Hughes, I'm going to cause trouble. Maybe not, I don't know. There was a clear signal, if you go back to people asked how OIX should work, and it was like a good Europe even exchange, non?profit. That's a clear signal that they don't want the exchange to be open to capture and therefore if you are going to go asking other parties to bid to run the exchanges, how can be be certain they are not open to capture.
SPEAKER: We are not asking anybody to do anything, we are basically saying why don't you make a proposal to do something and you get endorsement from from this community and that's how we prevent the capture by hitting these data points here.
AUDIENCE SPEAKER: It could be any company could put in a proposal to run an open IX platform. How are you certain that they are not going to be open to capture.
SPEAKER: Just the same way they are certain that the LINX or the AMS?IX. We have to have truth.
To expand on that. There is some trust, but on top of that, in the actual, the white paper that will release shortly, we do provide kind of a shoot horn in there to allow for influence over the governance of the organisation. Let's use the example of LINX. If LINX were to bid and were to run an exchange in New York or northern Virginia or somewhere like that, we need to make sure that that exchange has input from the people it directly impacts. If it's run by the large majority of the membership that is in the UK and Europe in general, as well as a board that is primarily in the UK, they are not going to have the same business interests in mind as the US. It's a very different landscape the and so as part of this, and when we publish the actual white paper, you'll see, we need to make sure that there is a way to have a structured amount of input from the open IX organisation into the ongoing operations of the IX.
AUDIENCE SPEAKER: Okay, so what you are trying to say, just let me get this right, is that the only people that you would entertain to run an OIX endorsed exchange are organisations which in themselves are not open to capture.
DAVID TEMKIN: Correct.
SPEAKER: There is some more bits in the white paper that you'll clarify and codify that and actually spells out that it should be a member onned kind of proposal and what not.
AUDIENCE SPEAKER: That's good. That needed clarification, because I looked at that and it was well here is organisations, they can bid to run this. Well how do you ensure that the clear message from your respondents to your survey is this shouldn't be open to capture. Thanks.
CHAIR: I think Joe was first.
AUDIENCE SPEAKER: Joe, just to follow?on Mike's comment, maybe a suggestion to do a renewal of the implementure, or whatever the heck you are calling it now, on an ongoing basis just as part and parcel of the whole process. And the slides, you had a minor error, you want the open IX was missing a dash, so some people might send mail to the wrong place.
AUDIENCE SPEAKER: We had the same feedback from Jabber.
DAVID TEMKIN: On the dash or on the renewal? It has a dash in it.
CHAIR: Thank you very much. I very much enjoyed it. I think you'll find a lot of support in the room for this style of organisation.
SPEAKER: Could we do a quick poll. Do people support the project and think we should continue to work on it? Do people think we should stop working on? Do people think we're insane? Awesome...
CHAIR: Thank you very much.
(Applause)
Before we go for coffee, closing for the first part of the EIX meeting a comment from Robert on some work that's being done in Atlas to do with the peering. Over to Robert.
Robert Kisteleki from the RIPE NCC ?? I am everywhere, it seems. Anyway, a bunch of people from the RIPE NCC, staff members and also some of you peering coordinators and others sat down on Tuesday for lunch, I believe, to talk about potential for using RIPE Atlas and/or RIS and/or other services from the RIPE NCC to help new your work in peering coordination, and IXPs and so on and I have been volunteered to say a word or two about that.
So we are going to get back to you on the mailing list to the Working Group to actually write down what we have discussed and to invite further input and just to see if this can be taken to a better place.
So, my job here is just to give you a heads up and you can expect some documents coming out of this discussion.
CHAIR: Thank you very much Robert. So, we're go to go for the break now. A few comments first of all. If you are interested in presenting at future EIX meetings, the call for papers is distributed via the EIX Working Group mailing list where you can sign up on the RIPE NCC website. This is where discussion about inter?connection topics, as well as details for the papers is sent. So whether you are interested or not, I'd recommend that you join in that mailing list.
Okay. We are about to head for coffee. After the break, EIX will be back in this room so there'll be another hour and a half of inter?connection topics before lunch, so hopefully see all of you after the break. Thank you.
(Coffee break)
