These are unedited transcripts and may contain errors.
EIX Working Group
11 a.m.
17 October 2013
BIJAL SHANGHANI: It's time to start the second half of the EIX Working Group. I just want to say we have got a new scribe, is Fergal and we have got Marco in the chat room, if you have any comments for people that are in the chat room, then Marco will look after you there.
We have one announcement. Tomorrow is the PC elections, so if you're interested in taking part in being part of the Programme Committee, then there is one position available so, please get your nomination in by 3:00pm today.
With that, we will start back on the second part of the agenda. And we have Kay, who is going to take about route servers.
KAY RECHTHIEN: Hi, I am Kay ECIX, I work there as network engineer and we had a few discussions in the last time in the office about routers, and I'm going to talk about it. The current situation is that we are a lot of routers on the the exchanges with different settings like some are filtered, some are completely opened, some are even on ?? if your router doesn't support it, then you have to do everything of that on your router. The problem is that if you have an AS set or the consents of the IRR DB, for example, if you take Hurricane Electric or something and you try to create prefix list out of that AS set then it gets really really big and then you have almost like 50 or 60,000 lines of configure and if you have even an up to date router which is really big for example and AMALIX 60, I think they end up at 50,000 lines of prefix filters so, you can't even filter every peer and routers service on your router any more. And other customers have the problem they don't even have an infrastructure or don't even know how to run automatic updates on their routers every night or every few hours. And others say if you have like RPKI and you want to filter base on that then you can only filter a PA announcement on RIPE and there is no real solution to PI address space and other RIRs doesn't even support it. So it's rather difficult situation if you want to have proper filters on your peerings.
So, the idea I would like to talk about and want to propose also with some other route server operators. Every exchange has a different policy so for example theey CIX, we run strict RPS filters for ten years, other exchange like the AMS?IX they don't have any filters because they say they are open so, it difference from every exchange. And the current route server, for example, we use birds, we have a really complex filtering scheme and you configure it out of a database or from the route server itself based on a /SKEUPT and you can basically insert every statement you want into the filter expression and even say like I want to have the current weather of the origin of the AS or the shoe size of the operator, and say I want to block if the shoe size is bigger than 45 in Europe, so you can filter on everything you want in really big and it's still really performed. And the idea is you had a BGP community base on that, so if the route is basically valid. In RPKI, you add a BGP community an you can do that the IRRDB. So the peer in that case which has a small how are or something which can't have big access list, just have to decide I want to have valid RPKI filters or valid IRR DB filters and just have to match waysed on one BGP community and reject everything else. So it's way for flexible and the customer can really choose what he wants to do, and so, this would be like example hard works, I don't even have bird code yet, but that's probably really easy so it's just some filter statements. So, you enter the route, it runs through all the filters you have like here in that case, it's IRR DB valid, attached to that community and if it's valid attach another community. Then the route with all the communities gets announced to the peer and the peer can just say in its route filters, filter on the community or else reject it or accept it or different local ?? etc. Than than having big filters every night on small routers. The thing I want to do is we should do it on the route servers, because if you take the ECIX currently around the biggest route server we have 120 peers around so the configure is around 100,000 lines long and if I put that like in a normal router like a Juniper that wouldn't even perform any more, so, you take the load away from the router and put it on the route server and you can make it more easier and more efficient to make all the filtering and you can make even more complex statements, basically you can just build a functioning RPSL filter completely of the AS numb object and filter based on that you can't really do that on a router any more. So, why should not use the capabilities of a current route server with all the aspects of filter and not wait for Windows till they write a support for RPKI or have really big filter lists that are efficient. So, you just make it more flexible than it is now.
That was just a thing I wanted to talk about. Maybe some exchanges implement it or think about it.
BIJAL SHANGHANI: Thanks, Kay, are there any question.
AUDIENCE SPEAKER: Will Hargrave. I'm interested in this work actually because we have our ?? realready filter on IRR on our route servers. I wondered if you had done any investigations like actually what proportion of routes on your route servers or routes on your route servers actually you know match these different categories or basically how much rubbish there is in there.
KAY RECHTHIEN: At the ECIX we have a lot of small customers and from my feeling I have to tell a lot of customers this they need a row object to announce something to our route servers and I just had a chat with somebody from AMS?IX and the Asian, they don't even even have an IRRDB, so there is a lot going away so that wouldn't next step to say let's implement that at an if you more exchanges and then see and analyse it like how much would be dropped if we matrixed filtering or how much on the exchange is currently RPKI valid so it's way more easier to get more detailed information about the peerage of the routes that are valid or invalid.
AUDIENCE SPEAKER: I think we should work on this.
AUDIENCE SPEAKER: Mike from AMS?IX. I was wondering how you got to the conclusion that AMS?IX doesn't do any filtering? Because we implement RRDB filtering as well.
SPEAKER: That was of the discussion for another company I work for, and as far as I know, the route servers are completely open, and you can ?? you announce ?? you filter based on IRRDB but only what goes to which peer and not RPSL based with route objects, but I can be wrong. I don't know what was the actual state now.
AUDIENCE SPEAKER: I just want to clarify, we are doing our DB policying so that members with filtering based on IRRDB objects, but we are not doing any filtering on the route servers per se. That's what I want to clarify. And to answer to Will, I think that the as built east people tend to not associate people with IRRDBs. The problem is if you get routes from Asia ISPs and stuff like that, those people do not confirm to the standards that we are used to in Europe and the US perhaps. So, what we see in the AMS?IX service is that about, let's say, 35,000 routes are theoretically not registered and would be considered collaborative, we just started tagging them. That was just a comment. No question, sorry.
BIJAL SHANGHANI: Thanks /AR on. Andy.
AUDIENCE SPEAKER: Andy bush, IIJ. The IR DBs suck just about everywhere except near DCix, AMS?IX, etc., which have a culture of using them. The only other place that the IRRs is at all useful is in Japan it's fairly accurate. In the States it sucks, Latin America, Africa, most of Asia except Japan, because they tend to be a little anal retentative over there.
SPEAKER: That's I didn't wanted to give the decision back to the customer so we are an exchange point and we are completely neutral but we want to help our customers as far as possible so they can decide on their own what they want to do.
BIJAL SHANGHANI: Any more questions? All right. Thank you.
(Applause)
Moving onto the next agenda, I assume we have not got peering personnels, so we have got three people that have submitted slides. If you are a network and you want to come up and introduce yourself, then we do have some time, so just if you can just queue on the side over here, that would be great. So we're going to start with Peter Losher from ISC.
PETER LOSHER: So, hello everyone, my name is Peter Losher, and I am with the Internet Systems Consortium. You know us obviously from the work that we have done in the past with regards to binding DHTP, you may also know us as in Anycast DNS operator. So, we have two Anycast DNS networks that we operate and I'll talk about here. One is F route, another is SN S, which are our secondary name service. We currently have for F?root over 50 locations worldwide and we have 17 in the RIPE region. And if your a network operator you are not already peering with us, we'd like to obviously peer. We work on the premise that we like to keep traffic as local as possible, so all of our global nodes are back in the US so if you are sending traffic to the US, we'd like you to stop, we'd like to you peer someplace nice and close to where you are.
We are not currently look to go add any sites but if there are any eye axes and so forth who would like an F?root note, get in contact with me. And if there is anyone here from GRNET or Gant, please contact me, because you are sending all your F?root traffic to Atlanta.
Our secondary name service, we have a public benefit instance where we host ccTLDs and non commercial zones, we have an Inns antidepressant of that in the rhyme region in Amsterdam. We have over 30 ccTLDs, including.nl and .FR, so that's ?? we are available over AMS?IX and NLix. So if you have any questions, I'll be here for the rest of the week. Thanks. Oh, and you can look at our peering DB page if you want more details.
BIJAL SHANGHANI: Thank you. Next up we have Kurtis from NetNod.
KURTIS LINDQVIST: I'm Kurtis. I still work for NetNod, even after the coffee break. And we also run a route server. And we also do Anycast for around 40, whatever, ccTLDs, and we are present at quite a few exchange points in Europe and in the rest of the world and also in some private locations, all the data is on the peering DB entry and the URL fell off the slide but the AS number is there, go look us up, we'll be happy to peer with you anywhere. Please do. Thank you.
BIJAL SHANGHANI: Thanks. Next we have Will van Gulik from IP MAX.
SPEAKER: Hi everyone, mime Will from IP max. You may have seen me around wandering and so on, so I'm quite a new guy here. I do host ?? manage six AS numbers. I do have L?root service interns, at 66 S BOP, so doing IPv6 things, you know and I do peer on these locations so if you are interested, just ping me. I'm around. Thanks.
BIJAL SHANGHANI: Next up we have Martin Levey From Hurricane Electric.
MARTIN LEVY: I just uploaded my slides 21 seconds ago, so... let's see if if works. I'll start.
AS 6939, everything is in peering DB as it should be. I think my biggest statement to anybody is please put your stuff in peering DB. We are available at an enormous number of exchanges in Europe, US and in Asia. We have an open peering policy. I think most people know that. And it's obviously v4 and v6 peering. So, I'm still talking waiting for the slide, but I think I have said everything I need to say and if anybody is interested, I'm going to make one last pitch for something. I have pushed for a few people to do NTU 9,000 peering, both on private N PIs and also over some of the exchanges that have capable of doing 9,000 or at least large MTU such as NetNod or the additional LAN that exists at DCix, N Y II X and a few other places. For those people that are interested we have done some pretty successful sort of cross continent, large MTU 9,000 testing from networks in northern Europe back out to California using multiple exchanges with routing through different paths and it does work, it works pretty clean, and if anybody else is interested in doing large MTU peering public or private, please track me down. There is a slide and I'm going to say thank you.
BIJAL SHANGHANI: Is there any other networks up here that want to come up and introduce yourselves? Nina.
NINA BARGISEN: Hi, I'm Nina Bargisen, I work for NetFlix, it's a video service operator, and currently we are present in Europe in UK, Ireland, Nordic countries and the Netherlands. It's AS 2906 if you would like to connect to us. Come and see me, I will basically taking care of the European part of the network. Our policy is open but obviously we mostly peer with people in the markets where we operate, but the future will come to the rest of Europe any ways. Thank you. AS 2906.
BIJAL SHANGHANI: Okay. That's the end of the peering persons.
Next up we have Nick, who is going to give a presentation on the proposal of the future of the EIX Working Group.
NICK HILLIARD: Hello everybody. My name is Nick Hilliard and I am CTO of INEX. And I want to talk about the EIX Working Group. Because actually, it's been a huge success. We have a huge amount of interest in the peering and inter?connection community, in a Working Group, in a forum which convenes at the RIPE meetings. And we have had two enormously important spinouts as of the European Internet Exchange Working Group as well, the Euro?IX forum and as a tangential to that, the European Peering Forum. There are not many Working Groups that can claim such an important legacy. But I think it's pretty clear that there are a lot of issues here that we need to deal with. There are a lot of content issues over the past couple of years that haven't been dealt with, and we tend to see a lot of same sort of things again, the IXP updates, you know, all that sort of thing, a lot of issues are relatively well solved.
In terms of work items, they are, for the most part, completed. To give you a little bit of history, the previous ?? when the EIX Working Group was originally constituted there were two parts to it. There was an open part and a closed part. The open part is what we have continued with today. The closed part hived off and became the Euro?IX Forum. And in fact the only form will al action item that we have left in the EIX Working Group is the IXP wish list.
It's two IXP centric. We have a vast diversity of people in this room. Some are ?? many are European. We have a large number of people from the United States, from Asia. We have consistently ?? we are seeing conditional more people from the African regions as well, and I think that this shows that there is a huge amount of interest in this as a forum.
And that we need to continue on with it, but in a more general forum. And what I would like to suggest is that instead of continuing on with the EURO?IX Working Group, that, in fact, we disolve it and that we create a new structure, an inter?connection BoF with the intention that if it's successful, which I strongly believe that it would be, that it would become a new inter?connection forum or inter?connection Working Group to meet at the RIPE meetings.
Obviously there is a little bit of bureacracy to be sorted out with this. There are going to be some organisers who are going to be needed to do this. And I think if we do this correctly, we could have a new and fresh and fibre rant community which can reconvene at the RIPE 68 meeting in Warsaw as a BoF.
And with that, I'd like to open this up to the floor. I think there are a lot of people who have a lot of ideas in the floor about how we can move forward, and I'd like to solicit some discussion on this.
AUDIENCE SPEAKER: This is Will Hargrave. I am an internet exchange operator and I support this proposal. We need to work and look at the success of the peering track at NANOG and how useful that is to the community and mirror that in the RIPE Meeting. Thank you.
AUDIENCE SPEAKER: Dave at the moment kin who up until last week was the Chair of the NANOG Programme Committee, the new term I'm not. I actually tweeted earlier in support of your proposal. It got filtered out of the tweet stream because there may be some people who don't like this but I am in full support of the your proposal as well. I have to say that you know, just to turn it slightly negative, I was disappointed in the pre?break programme here. I don't think it was a good use of our attendees time but I do think that there is a lot that people could do here. The ?? the repetitive slide equities of IX updates and giving us the same stuff that is available through marketing presentations is not really a productive use of this brilliant group of engineers' time. And so I think what you are proposing is a fantastic idea. It's time to move on. We have done a lot in NANOG to get ourselves to the point where the peering track is seen as truly useful. And it did flounder there for a bunch of years, but by and large for the last you know six, seven meetings we have gotten nothing but positive feedback and it's because we cut out things like IX updates. Some moderators choose to allow 30?second ones, and that's great. But there is a lot more to this than just telling us, as your famous presentation does, about up and to the right graphs and more ports. Thanks.
AUDIENCE SPEAKER: REMCO: I would strongly support your proposal and I agree with everything that David and Will have said. Thank you.
KURTIS LINDQVIST: I'd also like to agree with everything that David said and I think this is a great idea. I think it's time to declare success move on and try to refocus and the idea of having a BoF and outline a charter and topics would be a really good thing to do and a good way to move forward.
AUDIENCE SPEAKER: Christian Kaufman from /ABG mine, plus one.
AUDIENCE SPEAKER: Mike Blanche from Google, plus one.
Nick you have always been very good at shaking up this group, and indeed your famous presentation as Dave mentioned, that change this had forum somewhat over the past couple of years, there is certainly been some change but I think it is still time to move on.
AUDIENCE SPEAKER: I do believe there are networks that are not content networks that are also interested in this.
AUDIENCE SPEAKER: Ryan Donnelly from VeriSign. I'd also like to add my support.
AUDIENCE SPEAKER: Marco. RIPE NCC comment from Jabber: I support this proposal and I think it's the right way forward, inter?connection and IXPs are very important and highly relevant but like previous speakers we need more than just marketing updates and graphs, a clear charter and good content is needed.
AUDIENCE SPEAKER: From AMS?IX. I also support this proposal.
NICK HILLIARD: Okay.
AUDIENCE SPEAKER: Plus one from Linx as well by the way.
BIJAL SHANGHANI: Okay... there is one more comment on the Jabber.
AUDIENCE SPEAKER: Andy from IX ?? a Lego and former EIX Chair, I support a proposal to wider the scope of inter?connection topics at the RIPE meetings.
AUDIENCE SPEAKER: Mike Hughes, former EIX co?chair from a number of years ago. I completely support this. It's time to rechart this and do the right thing. I think the correct thing to do now, seeing as there has been so much support is to see if there is any dissent. If there is no dissent, declare a consensus.
BIJAL SHANGHANI: Is there anyone in this room, the Working Group, who does not agree with closing down this Working Group?
AUDIENCE SPEAKER: Maybe, Nick, could you just at this point it would be a good point to recap what you have to propose so that everyone is clear?
BIJAL SHANGHANI: Good idea.
NICK HILLIARD: Okay. So, the proposal, then, is the disillusion, the formal disillusion of the EIX Working Group; to create a new BoF to reflect the values of general inter?connection rather than the specific focus that we have at the moment on European IXPs, and if this motion is carried, we need volunteers from the floor for organisers to create the new charter and if we have that, then hopefully we can reconvene at RIPE 68 with a new charter and fresh ideas and I would hope a much more vibrant community.
AUDIENCE SPEAKER: Martin Levey. Can I ask for some clarification on one item? We presently get half a day, essentially, of session which is provided on a video stream. If this change occurs and it becomes a BoF, what guarantee do we have of time during the RIPE Meeting and what capability for remote participation would exist?
NICK HILLIARD: Okay. So, the process to create a new Working Group is to create a BoF first, which will meet on either the Monday or the Tuesday night at the RIPE 68 meeting, and that will include all of the standard remote participation options. I see Rob heading towards the Chair ?? microphone ?? towards the Chair ?? and at the BoF, I would expect that if the BoF is successful, that it will formally request the to the Chair of RIPE that a new Working Group be created, an inter?connection Working Group.
ROB BLOKZIJL: On a personal note in the first place I fully support this proposal. In my role as Chairman, I think it is an out?of?the?ordinary procedure in the sense we have never had it before, a Working Group has said we close by we start a new Working Group. So, I think the way forward with organising a BoF, finding organisers for that BoF is a very practical way. And I will check with the RIPE NCC that the facilities that you need will be there.
NICK HILLIARD: Thank you.
AUDIENCE SPEAKER: Dave Wilson, acknowledging the unavoidable problems with Fergus not being able to get here, and thanking by /SKWRAL for running the session. Does anyone know what the Working Group Chair thinks of this?
NICK HILLIARD: I think we have been having severe difficulty in contacting the Working Group Chair over the last 24 hours. This is not due to not for want of trying.
BIJAL SHANGHANI: I did get an e?mail from Fergus and he does agree that the charter does need to be ?? the scope of the charter of the Working Group does need to be changed, so, I think that ?? well I can't speak for him, but I think when he sees the support that the Working Group has on actually you know dissolving this and starting something new, which is you know of course for the community, I think ?? I don't think he'll have a problem but he is going to be here later on. He is due to arrive around lunch time. So, there is an opportunity to speak to him then.
AUDIENCE SPEAKER: Thank you very much.
BIJAL SHANGHANI: So, is there anyone in the Working Group that believes that this Working Group should not be dissolved? Just for the audio or video, there are no hands.
Well I think there is consensus then, and I think we can ?? Kurtis?
AUDIENCE SPEAKER: Kurtis, you should probably ask the other question too: Who is in favour?
BIJAL SHANGHANI: All right. Who is in favour of dissolving this Working Group? Have a show of hands or a collapse or hums or everything.
(Applause)
In which case I think there is consensus and I think we'll move forward and dissolve this Working Group. Rob, do you want to ?? is there anything else you want to add?on that?
ROB BLOKZIJL: I just want to repeat what some of the previous speakers already said. This is a Working Group that is dissolving itself not because they don't know what to do any more, but because they, as Nick so adequately said in his introduction, has spawned a lot of successful results and is now ?? it's now a matter of reorienting, recharting and the easiest way to do is dissolve it and start from a fresh start.
A couple of administrative points on the Friday morning, Closing Plenary, I expect a short report to the community at large that this Working Group has now been dissolved, but if efforts are on the way to have a rechartered Working Group and the first step will be a BoF in Warsaw is one thing.
Second thing is it would be nice if by Friday a coordinator or organiser of that BoF can be announced so that people who are interested know who to contact.
BIJAL SHANGHANI: Thank you. Marco.
AUDIENCE SPEAKER: Andy Davidson on Jabber. If any inter?connection Working Group is formed please can all subscribers from EIX be added to that group list.
BIJAL SHANGHANI: Sure. Okay. With that, I think consensus has been reached and we dissolve the Working Group. So, with that...
(Applause)
With that, I would like to thank the chairs. We know that Andy is online watching so I think that Andy and Fergus, best of them, they have done a great job with the Working Group, and you know, as Nick said earlier from that we have managed to get EURO?IX and the EPF and those are both very successful forums. So, thank you Andy and Fergus for doing a great job.
(Applause)
AUDIENCE SPEAKER: Will: I'd like to thank Bijal for stepping up at rather short notice with Fergus as travel problems and being an awesome Chair, thank you.
BIJAL SHANGHANI: Okay. Next steps. What are we going to do? Right, like Nick said we need to have some kind of BoF for RIPE 68, so, we, as mentioned earlier we are looking for two interim chairs, so are there any volunteers right now that would be interested in chairing the inter?connection BoF at RIPE 68.
AUDIENCE SPEAKER: I would be interested in working with Nina on this project.
AUDIENCE SPEAKER: Nina. I'm Nina, I'll volunteer. We have got at least two volunteers and one more person at the mike possibly.
AUDIENCE SPEAKER: Martin Levey. Yes.
BIJAL SHANGHANI: Okay. So we have three excellent volunteers, so, with that, I think we can discuss ?? oh, Mauro from MIX.
AUDIENCE SPEAKER: Mauro. I would like to participate also.
BIJAL SHANGHANI: All right. So that's four volunteers already, so what we'll do is I think the four of you should get together during lunch and ?? perhaps with Nick as well, and decide the best way forward on this, on who can Chair the inter?connection BoF at RIPE 68.
Mike?
AUDIENCE SPEAKER: Mike. Point of order, the BoF organisers, what we should do obviously is the first thing they should be doing at the next BoF is working with the community to define what the charter should be, and see how that pans out in the first BoF. And then they can fine tune the charter and then we can request the Plenary to become a Working Group and at that point we can then basically decide whether, as a group, whether they want to continue as chairs. They may not want to continue as chairs having done this, so, whether they want to continue as chairs and who shall do that. That's just a point of order about how we do this process I think. Thanks.
BIJAL SHANGHANI: It's Nina Bargisen, Ren Provo, and Mauro Magrassi and Martin Levey.
Right, are there any more comments on this topic? No...
Okay.
NICK HILLIARD: Thank you very much everybody.
(Applause)
BIJAL SHANGHANI: All right. And with that we move on to AOB. Is there any other business to be discussed?
All right. I think we're done. I don't know what the time is, but I guess ?? I think we are finished early, so we'll have a little bit longer for lunch. And this is the last time we'll be here at the EIX Working Group, it's a little bit emotional, I am not even the Working Group Chair but I'm feeling a bit emotional, but you know we have got this new fantastic new idea for the inter?connection BoF and I'm sure that's going to be a success. So we'll see you all at RIPE 68 in the inter?connection BoF. Thank you.
(Applause)
