Connect Working Group
30th November 2023
At 11 a.m:
WILL VAN GULIK: Hello, everyone. I think we are almost on time. We have got a minute to head to your seat fasten your seatbelts.
FLORENCE LAVROFF: Welcome, everybody, we are about to take off.
WILL VAN GULIK: Welcome to your second favourite Working Group, the first being RIPE NCC Services, obviously. So some important information, we have two emergency exits at the right, on your left, and I think two or three at the right, so be aware of that. No lines on the ground, that's okay.
Welcome to connect. So we have got quite a packed agenda for this session, and so, first of all, we would like to appoint our scribe, Alain, thank you very much. Our marvellous agenda, and we know that you all went to the RIPE, at the last session minutes that have been published on the mailing list. If anyone has any comment about that? I don't see anyone reacting.
FLORENCE LAVROFF: We can just move on to the next topic then.
WILL VAN GULIK: I think, I think so. So we had some ‑‑ we were reminded that our process for the selection of the Chair, election process was like maybe a bit unclear, so we will ‑‑ oh, yes, that's also true ‑‑ before that, Connect is almost ten years, can you believe that? So it was just a short reminder to give you all the information about what was actually our mission, so you can have a look on the website and, well, think about what our mission is still happening or if we are doing still doing a good job for that and if that fits our mission.
So with that, obviously we need to clarify the selection process. So, that's not how it works. We discussed at the last time at RIPE 75, so before I was a Chair, so the process ‑‑ the selection process goes like this:
So there is a call that's made to the mailing list every three years, or whenever needed; the interested parties have two weeks to make their interest known via the mailing list or directly to the Chair; after the two weeks, the Chair ensures that all the candidates are announced on the mailing list and issue a call for discussion; the Working Group members express their approval or otherwise of the presented candidate; and two weeks after that, after the call for discussion, the Chair declares a decision based on the mailing list discussion as they would do for a policy proposal.
So, this is our call, because we didn't feel that it was needed before, and so, if anyone is interested or curious about what is our mission here as chair and if they would like to participate, you are all more than welcome to talk to us or to mention your desire on the mailing list.
FLORENCE LAVROFF: Please reach out.
WILL VAN GULIK: Yes. And don't hesitate, we can explain you what we are actually doing here, but you might have an idea.
With this, I think we can move to our first speaker.
MIRJAM KUEHNE: I am RIPE Chair, I have one question because you said something about RIPE 75. So three‑year terms, so ‑‑ if I do my math correctly, there should have been one last time or this time?
WILL VAN GULIK: That's the point we said, we make a call on the mailing list every three years or whenever we feel it's needed or whenever needed, so basically, no one expressed any interest, we usually go to the Working Group and remind everyone that we can come to us if they've any questions or if they want to be part of the Chairs for this Working Group, so ‑‑ but that's why we are making the call here because obviously we didn't have any feeling about that or felt that it was needed so now it's done and so we will try to remind that every three years because ‑‑
MIRJAM KUEHNE: This is the open call for volunteers?
WILL VAN GULIK: Yes, this is the call for volunteers.
MIRJAM KUEHNE: You might want to repeat that on the list after the meeting
WILL VAN GULIK: Certainly, that was our target.
FLORENCE LAVROFF: We do not wait for three years for people to manifest their interest, people can just jump in when they want, contact us directly on the mailing lists, we are there.
MIRJAM KUEHNE: I think it's good to remind people from time to time to there's this opportunity because otherwise they don't know that there's space because nobody is stepping down regularly.
WILL VAN GULIK: Absolutely, that's totally agreed and that's a good information for newcomers also.
MIRJAM KUEHNE: Exactly.
WILL VAN GULIK: That's also an opportunity for us to clarify that and to update the website with the policy and with the process for this selection.
MIRJAM KUEHNE: That's great. Because I think people expect some term limit or some clear guidance and ‑‑
WILL VAN GULIK: That's why we are doing that so thank you for nudging us and now we have a clear process and information and we will get back to that regularly.
MIRJAM KUEHNE: Thanks.
WILL VAN GULIK: Thank you very much. Okay, and so, I think with that, we can go to our first speaker.
FLORENCE LAVROFF: Marco from Open‑IX, please, join the stage.
MARCO PAESANI: Hi, everybody. Ladies and gentlemen, good morning. I am really pleasure to be here with you today. But to be honest, I wasn't the one to talk about this important subject, I decided anyhow to try and share with you a few ideas, and I hope you will see through my eyes how I see Open‑IX in the present day in the connection landscape.
I was born in Rome and grow in Rome but I spend more time to dedicated to living in Florence. You know Florence? Yes, everybody know Florence. And I did think not this is just a coincidence. As you may know, Florence is the ‑‑ is where the Renaissance was born and when you are thinking about Renaissance you meet feel lost in a forest of artistic notions, painting, sculpture, as historical when all artistic expression sculpture, paintings, building, literature, and reflect the complete and total overview of the previous social standard because today we talk about standard.
First of all, we discuss about you, about Giotto ‑‑ let's start with my first slide and consider art the failure of Renaissance, we worked in Florence between 1200 and beginning of 1300, the town has just reached 90,000 people. But it's landed at 40,000 people half a century before. This means Florence is growing extremely quick and reach the same scale of London and Paris, three times Rome but don't tell Roman people because Rome is not bigger.
Places to meet, to stay together and to connect.
Do you know how and when, how connect each other in those days? Going to the church, attending the mass, praying together. Churches were the ‑‑ there were more than a dozen new churches under construction in Florence. Giotto was the great entry of the churches, do you think that satisfy with the time treaty standard? No way.
Because standard is so difficult, in a place Florence was richest but built by man, Giotto was interpreted in a man but he had a times standard putting under the spotlight ‑‑ normal as a result here is Mary is the Mary in majestic, is bigger than the angel and this was normal to highlight the divine personage in painting and baby Jesus is not baby, but growing up 30 years old man. This age Jesus had started to preach.
On the right ‑‑ sorry, but adopt this rules and decide to add rose on Mary's cheeks and her lips as you can see, visible under the dresses and that human detail never considered important before, representation standards started to get changed legend.
On the right side of this slide you can find the picture from Gentile de Fabriano. 100 years later the challenger was still alive. New element was considered old, mystic and so far from the ‑‑ to everyday life. People addressed the adoption to Maji, reflected not just time fashion but 1400 fashion in the town where a rich man like ‑ can be proud of a collection of Arabian horses and painting commissioned to Gentile De Fabriano, no ‑‑
Masaccio is on the left. This was painted, another younger painter, he was trying to ‑‑ and challenged every way the holy Mary and Jesus and the painting till then. He could not change and get more human the expression of Mary and the Jesus or the angel faces is Saint Anna. Faces looked serious and stiff was normal in those days but he adopted a new standard for body proportion, light now he hit from the specific source and body appearance, for the first time Jesus is not covered with dresses hiding the body but in a real human body with athletic hearts.
The next picture is by Filipo Lippi, do you think this was enough? No way. Mary and Jesus had become really ladies and babies, but to get to this point a monk in love with a nun. Scandalous story around the middle of the 1400s, the love affair between them was completed out of the rules but it was real ‑‑ the love affair was completely out of the rules, of family and fill the life of both of them. That's why Filippo painted here in this other painting.
She came inspired ‑‑ first life and have ladies' face to represent the holy Mary. Another great painter, Sandro
Botticelli as you can see now, but he was not satisfied with his master's standards, he went on, looking for a new criteria to express his sense of life. He didn't want to just paint most beautiful Floretine ladies, he wanted to hit those ladies here significance and Greek mythology, human figures, his painting and moving gently and elegantly by ballerinas, driving by love, longing for love and surrounded by famous nature ‑‑ perfect. Spring and the birth of Venus two famous paintings for the number of visitors in Italy who visit gallery. ‑‑ of constant research and new criteria to get right and older and get deeper perception of the life.
This great man with many other Florence's art history are just a little taste of what we found in the human history, a constant need to challenge with what is given by us the previous generation and just to develop better landscape. When the communication and connection with the requests and the involving the world.
For this reason, there is standard in our market also. And Open‑IX is a standard, it's created a global standard with three different certification, OIX 1, OIX 2, data centre, and OIX 3, edge data centre.
This standard we cover all our job, about connection and it's most important like we have a standard for your data centre like tier 4, tier 3, this is new standard tier in the and it's most important to be standard because you have a new way to connection each other. . An example: IX. This standard encouraged IX to penetrate a new market in geography because it's standard but today there is a no standard. 1 IX is built in some way, the other in other way. Many IXPs are not for profit and for awe truistic reasons, such as expansion of Internet services and reach, as well as for improved latency for new services. For these specialised Open‑IX offers free certification, it's most important.
Data centres as certification encourages to penetrate also the same market and also for data centres, if no profit, also for this one he have the same certification.
About the community, because Open‑IX is an association, is not company, and ‑‑ in addition to maintaining standards OIX provides actionable information and intelligence about the interconnection landscape. The first product is the interconnection navigator which provides an interface for research and visualising interconnection data getting from Peering DB. An example: This is how the OIX grow in Italy, years by years. All information is contained in database of Peering DB but with this interface, you can do much better and much faster. You can scan the QR code if you want to immediately.
I think is enough from me. I am sorry for a little delay, if you learn more about Open‑IX I am here today and tomorrow. Any questions?
FLORENCE LAVROFF: Thank you, Marco, for those well ‑‑ for this well needed arty moment. I hope it will make for the ugly font of the past slides presented by the Working Group before, pink Barbie is the colour of the day but I am not sure that's going to make up for it. Anyway, do we have some questions from ‑‑ online? No, thank you, Anand. Do we have a question from the floor and I can see Arnold approaching.
ARNOLD: From deduction and association of internet exchange in Asia. My first question: How does OIX align with the IX‑F that we see, association of internet exchanges all over the world and we still like to have an association of internet exchanges in North America, the IX‑F also has its definition of what an internet exchange is.
Second question is also for data centres, there are much certificates and standards around for data centres, how does OIX align with these standards?
MARCO PAESANI: Thanks for the question. About the first question, probably I am not right people because I am only in the marketing side of Open‑IX. And I will give you an answer when I come back for sure.
About the second question, and the standard about what you mean is so different from our standard because the standard of Open‑IX is relative to the people working inside this data centre, this IX, is very specific standard, is not the same standard. Is not talking the same argument. And you can check it out about this difference between making cooperation between the standard, an example tier 4 and Open‑IX PC is not the same standard, is not the same goals. We have a different target. The first one is about the structure, the other one about the complexity of the whole information. This is the main difference between the standard. From my point of view.
FLORENCE LAVROFF: No further question? I think that's it, we can move to the next topic. Thank you, Marco.
So the next topic is going to be online talk from our fellow co‑chair Remco.
REMCO VAN MOOK: Can you hear me?
FLORENCE LAVROFF: We can, Remco, so the topic is about adoption of Address Policy for IXPs, the on‑line floor is yours.
REMCO VAN MOOK: My apologies for not being actually in the room with you, I got the dreaded two stripes on my Covid test this morning so I thought I had better not infect you guys. I am going to talk really quickly about the policy proposal that would reduce the default assignment for new IXs from 24 to a 26 that has been discussed in Address Policy, and was adopted earlier this year after a previous RIPE meeting.
So, I thought I just wanted to make sure that we address that here as well, new policy has been fully implemented so anyone who is coming around with a new IX right now will have to deal with default assignment size of/26 from now on and that's really it.
WILL VAN GULIK: Okay, perfect. Do we have any question or comment about that? Thank you very much, Remco, so now we can move on with our next topic, I guess, and we have the follow‑up on the discussion we had last time about the proposal for common use of the IRRd B.
STAVROS KONSTANTARAS: Good morning. It's time to give you some update regarding our document, our proposal that we did in the previous RIPE meeting here in Connect Working Group, about the use of IRR databases in the IXPs and especially in the root servers.
So, quick recap what is this document about, this proposal is about IXP operators to trust and use only the following IRR databases in order to configure and maintain the filters for the root servers, in the previous document as you can see in the previous proposal we came up with this list of the big five IRR databases and also, we noticed and we mentioned that, we understand this adoption of this document will probably result in a massive transfer of RPSL objects from unsupported databases to the supported ones and therefore the policy also introduced 12 months grace period where actually the big five IRR databases are extended a little bit more, the list is extended a little bit more with the following four non‑authorised databases so we can allow the people, we give time to the people to go and be able to do those transfers to the big databases, officials databases.
Also, we did some work ‑‑ we didn't mention a lot of things in the mailing list but we did a lot of work in the background, the authors, and first of all, the first change we did was about should be a policy or BCOP, we had this internal discussion. We feel it is more of a BCOP as policies have this punishment, let's say, close, that if you don't follow you are a little bit punished or there are some consequences, in our case there are no consequences so it's more of a best current practice for IXPs so we decided to change it from policy to BCOP or actually a strong recommendation. If you guys think this is not a good idea of course feel free to come up with your suggestion, but that's how we understand that it should be this document.
The second update that we did as the authors was that we understood that some of the big IRRS have national IRR delegations so the document have already explicitly mentioned in the website that okay, we host these amount of objects but if you go into the national IRRs that host the other objects, so we felt that this, it's better if we include those NIRs inside the official delegations so the list has been extended, for example, LACNIC has officially authorised NIC dot MX, NIC.br for Brazil region so it is fair to include that database inside our list since it is official delegation.
The next change that we did is that okay, initial proposal had generally first 2024 as starting date, we decided to move it and the reason was we want to have at least two RIPE meetings that we can discuss this document with the community, come up with ideas, proposals, everybody is aware, and that means that okay, starting from this RIPE meeting we are going to have the next RIPE meeting for discussions, so the current proposal has June 1st 2024 as starting date. But okay, if the whole room says oh, it's a bad idea, we don't like it, we should postpone it, we are free and would like to hear your feedback on that, freely free to come to the microphone or mailing list and discuss it.
Of course during our investigations we had a lot of work in the background, we had some unsolved problems like for example famous tier 1 provider that has legacy block and small /24 blocks to clients and then the lease holder cannot register his /24 to the database of ARIN and then ‑‑ but this customer wants to connect to European based IXP and announced this /24 space and then announcement of course BGP announcement of this space goes to the IXP service that means if I build a prefix list from ARIN database to accept this /24 it will not be there, it will be an empty prefix list.
So, we don't have a solution for that, of course it's a corner case, we don't see a lot of those use cases but they do exist and we would like to invite the community if you guys have already thought of some ideas or you already have brainstorm or have clues how to tackle those problems we welcome you to provide your feedback, perhaps in the mailing list and you come to the authors and tell your ideas and we can consider them. For now for us is an unsolved problem within the boundaries of this document.
Of course there is shoe that fit all sizes, right? There's no perfect draft that can solve every potential corner case, use case that exists outside. The Internet is a little bit ‑‑ but that means, however, we might need to sacrifice 1% of prefixes if we need to save the other 99, we feel it should be our approach, our strategy. We had some questions in the past about why don't we do RPKI and we work on that but this document actually helps adoption of RPKI; we feel that is a pre‑requisite and adopt RPKI faster so we don't see this work as colliding with RPKI work but as a parallel way, right? So we need to continue working on that document and that will spread adoption of RPKI even with faster rate.
Also, we discovered and we think RIPE ‑‑ thank RIPE for the information. We see that RIPE is also able to bring to the space, to the services, if you have a block that belongs to the legacy space and you want to bring it to the RIPE region and use it in RIPE region RIPE will accept and put it in the registries and that means that you can find potentially this space in the RIPE database and in the RIPE services.
So, that's very good for us, that's very positive. And during that period we also talked to a lot of IXPs, also the big ones like DE‑CIX, LINX, CZ.NIC, they like the idea and think we do valid work, they think this is also spread the adoption of RPKI so they are want to work on that with us and support the idea so that's very positive.
And actually, we decided with the DE‑CIX people and also with Markey from MeNap that we are going to work together to finding methodology that we can measure the impact of the adoption of this work so we are going to measure the impact and hopefully in the next RIPE meeting we can share the results with you.
So far, we have got very, very good support from the community on that.
And that concludes my small update and I welcome your questions and comments.
FLORENCE LAVROFF: Before that, do we have questions online, Anand? All right. Let's start with questions from the floor.
WILL VAN GULIK: Policy writer or document writer. In this case also, I have additional information because I almost resolved the case of the 44/9 space so now I will be talking or trying to put them on the mailing list to say okay people, you need to have your prefixes and your route objects created in the ARIN database or somewhere else or you need to think about fixing that but I think I found a way to go and reach them so that slide one almost /8 considered and so that's good, positive improvement. So we will see and we will get back to you on that one and that information came yesterday so why we didn't include that in our slides.
STAVROS KONSTANTARAS: Work in progress.
BLAKE WILLIS: Blake from Zayo. Thanks a lot for this, it's definitely something that I personally support just to get the stuff cleaned up. Have you had any or anyone in the ‑‑ any of the policy writers have any communication or feedback with the RIDB people about this because there's still specifically in North America there's still a lot of ISPs that use that as their like de facto registration?
STAVROS KONSTANTARAS: There is a lot of de facto registration or usage of that but we do understand that there's a big portion of this usability that comes back of ‑ and not because there's an actual problem right because it's easy to register and easy to use, they don't want to bother ‑‑
BLAKE WILLIS: Mostly for legacy reasons, we always did it that way.
STAVROS KONSTANTARAS: Easiness. But there are of course problems right? We already mentioned this case with tier 1 and then perhaps those people go to register their objects over there because they need to be somewhere. This is again an unsolved problem but this is the smallest portion of the case, the majority is not this corner case
Blake: The question has been there been communication with the RIDB people, are they aware of this, that the root servers will ‑‑
STAVROS KONSTANTARAS: I don't think so. We didn't communicate on that. I think however they are aware. Somebody mentioned to me something that ‑ solve this new document and I think that's why they moved very quickly to D4 because they saw this document, however they are not going to delete objects, they are going to mark the RPKI invalids and that's it, the only thing they will do.
WILL VAN GULIK: I can comment on that, I think I will try to contact them and say okay, by the way, we are doing this, do you have any view? So I will take that and I will do that for this matter.
SPEAKER: A partial solution to the issue of US legacy space holders is that somebody could set up a new IRR or maybe a section of an existing IRR which only accepts route objects for legacy space. This will at least avoid the polluting the rest of the Internet and limit the problem of fake objects only to legacy IP space but somebody needs to volunteer for that, I think. Marco, document contributor.
STAVROS KONSTANTARAS: Which is a solution we thought yesterday at 1 a.m, right.
MATTHIAS WICHTLHUBER: Matthias Wichtlhuber from DE‑CIX. Thank you for moving on from this proposal, I think this is really important. I have done an upfront analysis for our customer space and I see the IRD B issue so there are exponentially more peers than for other or for all other databases together that are alternative so I would like to propose to be very careful with the fade‑out, maybe give it more time because we have experience with numbering peering LANs and that's not fun
STAVROS KONSTANTARAS: It's not.
SPEAKER: And it can take a lot of time, actually. Thank you.
STAVROS KONSTANTARAS: Thank you, Matthias.
SPEAKER: From Juniper networks. Just one small remarks because I have been working in the past with various toolings and I would say we always speak about what to trust regarding data sources which is great first step, we should also think of how we use that information, that means the existing tools, most people are still relying on old scripts that are using that IS set one which expands that with least of recursive expense to the least of AS numbers, and by the way using also BGP Q4, BGP Q3, scripts and all of them are using IRRd B so the point is what to do with those tools so should she contact all databases exponentially or as an alternative, is there any initiative from the community to create a central repository where all trusted databases would be mirrored, like the new IRRd B that would only include trusted sources?
STAVROS KONSTANTARAS: The thing they are two different answers that I need to give you on this long question, regarding the two ‑‑ I know for BGP Q4 by default contacts NTT.com with a minus H option you can say go to this database and bring the objects from there, and if you go to a large database or mirror database you can specify with minus capital T option the short and they can say bring me this with object with RIPE source or ARIN so you can specify the sours. The defaults are always a question of course and perhaps that could be a nice thing to do to update the defaults on all those tools but that's a next step, let's say.
And regarding mirroring, I know RIPE mirrors all the other databases every day at midnight so it is possible if you contact the Whois service of RIPE, you can get mirrored objects from ARIN, APNIC and so on. So, if somebody wants you can have RIPE as one‑stop‑shop to do that.
But you know there is always a 24 hours delay because they do it by midnight and if you are going to be faster you have to go to the individual ones. I don't know if that answers your question?
SPEAKER: Yes, thanks.
Will: Just an extension on that, we discussed the default of many softwares and we said maybe we can ask some of the software developer to say it should not be ready DR those kind of stuff so we will also push in that sense.
STAVROS KONSTANTARAS: I think we can make a list later on, go in the mailing list and create a list of those tools and the defaults and when this document has adoption we can take it as a next step, yeah.
FLORENCE LAVROFF: Awesome, thanks a lot. Any other question? It doesn't seem likely, thank you.
STAVROS KONSTANTARAS: Thank you and sorry for all this time we consumed.
FLORENCE LAVROFF: Thanks great, thank you, probably something we want to continue on mailing list.
Our next speaker is Leo from PeeringDB.
LEO VEGODA: Hello everyone. So, I have an update for you from PeeringDB. So, the focus will, to some extent, by on campuses and carriers, these are the two new objects or data pilars that we introduced at the start of this year so I want to take you through what we have been doing with those. There will be some updates on other stuff but before we get there, I wanted to let you know PeeringDB is mostly volunteer‑run organisation and you have the opportunity to both volunteer and vote. There's a new how to on how to make sure that you are a member, if you are in this room there's a good chance that the organisation you work for could be a member, and this will tell you how to be a member, how to vote, we have just had new people join the admin committee, the product committee, should hopefully behaving some new people join at the end of this year or early next year and there will be an election as there is every year for the Board, sort of earlyish next year so take a look at that how to. If your organisation is not a member and there is no one there to vote yet, then it's your opportunity to influence PeeringDB. There are people in this room who are on the Board of PeeringDB and Bijal was recently on the Board and left PeeringDB and if you have questions, look up the Board members of PeeringDB and ask them if you are interested in running.
So, I mentioned the new objects that we added, the campus was one of them, we created it in February. We weren't entirely sure at first whether people would want it and so we didn't put it in the main API but we added it in August, we've put in some other UX improvements so for instance where something is a campus you will see that little multi building sign next to a facility to show it's a facility that's part of a campus, we have added it to search in October and literally in Beta that was released yesterday, into Beta, you will now be able to see all the resources for a campus in one view, so that you can then go and filter for each kind of resource that you might be interested in and find out if there's something there that is useful to you. Please go and take a look at that on Beta.PeeringDB.com, it will hopefully be released into production next week.
We have done something very similar for carrier, people liked it so we automated it, added it to the API. We have put in some other UX improvements and it's now visible in campus, so you know we have been continuing to improve that throughout the year, it's now coming close to fully supported, there are still some improvements to be made but you know, we've really been trying to make these fully featured, features.
So, V2 search, this is meant to be a better search that is both better operationally but also for you as users. We introduced it in May and asked for some feedback. We got a little bit of feedback. We alphabetised the results in June and then put it into the default position, now we are starting to get some more feedback which is genuinely useful. We still have two or three bugs to fix and then at some point we will remove V2 search ‑‑ we will remove the old search when V2 doesn't have the bugs that people are reporting.
An example of one of the bugs that we will be fixing is at the moment in V2 search you can't search for a peering LAN IP address, so we know about that, it's going to be fixed. But please go and test it and use the form to report bugs if you find them because that's how we make fixes and make improvements.
Web interface, this is one that we know is divisive, some people really like the simplicity of the site; other people were saying it needs to be better. So, this is an example, one of the designs for improvement to the UI, Rebecca has been working on the design for us, sort of keeping it the same but making it sharper and more responsive. We will be adding some personalisation features so that, for instance, when you log in, you can go and see the thing that you care about and not worry about other stuff. And once we have done all of the changes to the design, we will do some lazy loading which should improve the user experience, though it's obviously not going to actually make data arrive faster, but you know, we want anyone who is using the web interface to have the best experience that they can.
We just released updates from a source of truth, this is something that was developed initially at NANOG hackathon a couple of NANOGs ago, it's now in production. The idea sufficient some sort of internal source of truth, it can go and scan the differences between what it knows and what it sees in PeeringDB. It can then send a suggestion to PeeringDB and a human can go and log in and either approve or reject those changes. This should hopefully improve the quality of data in PeeringDB and therefore be a win for everyone. It also means that you don't need to give your internal source of truth an API key to go and push changes that happen without human supervision, so this is good for people that have lots of entries but maybe don't have full automation. So, we hope you like it and if you need anything on this then obviously let us know.
Got a couple of quick updates with updating the PeeringDB client, doing a full rewrite on that. We will be closing PeeringDB Whois at the end of January. It's just another interface to the API. It needs a full rewrite. That would be expensive. We don't really think very many people use it so this is an opportunity for anyone who absolutely relies on it to let us know why they rely on it because it's expensive to go and make it work the way you will want it to work. You can link your social media accounts in PeeringDB if that's important to you and we now have all of the facility data in PeeringDB published as a KMZ file, published every day. We already have suggestions for improvements to this. Some people would like continent‑specific files, there's an idea to make the advanced search exportable in that format; other things, if this is useful to you then please let us know how we can improve it and make it more useful.
One thing to note is, if you want all of the data that you see there, you do need to use Google Earth pro or something like that, the pro)web‑based things don't show all the data.
Thank you very much to all of our sponsors. We couldn't do any of this without our sponsors. And if I hand back to you.
FLORENCE LAVROFF: Thanks a lot, Leo.
No question from the chat. So we will start with personal questions.
BLAKE WILLIS: Blake from Lazayo: Thanks a lot for this, really good stuff, the KMZ in particular is really cool, I thought people will be all over that in about five minutes. Specifically the carrier object for carriers like us is really useful, I would encourage anybody in the room that has lots of ‑ buildings to use that instead of responding to the enlist query of, endlesss queue of data centre operators asking you to add their buildings to PeeringDB so it looks like you peer in thousands and thousands and thousands of locations, the carrier object is super duper for that.
SPEAKER: From DE‑CIX and also a member of the PC at PeeringDB and I was also on the AC from PeeringDB so I have a question.
So, if you look at the GitHub of PeeringDB there are tonnes of issues, not resolved, some of them hanging around, is there something which the community could do to speed this up that gets resolved, some of them for example, we had this Working Group on data ownership which is not yet concluded, we have IXs where you can import information from which is matched with the information from networks. This is implemented halfway but not full way. The others is, when a network has been set up we get information from the RIRs but then it's not never checked again. That means if an ASN is returned back to the RIR, PeeringDB does not yet check. There's also GitHub issues, both of them would massively increase the data quality of PeeringDB. Is there anything the community can do to speed this up?
LEO VEGODA: I think the key thing that the community can do is speak with people like you on the product committee or me as the product manager and let us know what your priorities are and why they are your priorities and the impact that those changes will have. Because the more you communicate the impact, the easier it is for the product committee to resolve the issues that are still in the decide status and then we can move them forward. That said, we are able to resolve sort of like 70 to 100 issues a year, depending on their size. We have about 100 issues ready to be implemented now, so we have made progress, it was like 130, so, we have reduced the size of the backlog, we are making good progress but communication with the community and between the community and the product committee, if something would make a real difference to your business or the way that you do your work, please let us know.
SPEAKER: Okay, thank you, Leo, and thanks also for the work that everyone does at PeeringDB, I really think it's a great resource.
FLORENCE LAVROFF: Very true, thank you, Arnold. Anyone else? I don't see anyone. So, thanks a lot, Leo.
LEO VEGODA: Thank you very much.
FLORENCE LAVROFF: Let's move on to our next speaker, Bijal from Euro‑IX for our traditional Euro‑IX update. Thank you, Bijal.
BIJAL SANGHANI: Hello, and thank you, Florence and Will.
So I am going to give you an update on Euro‑IX and the way that, the format of this is I will give you an update of some of the news from the membership and then a little update on Euro‑IX and today I am going to do something different; I am going to have an interactive session using Slido so on some of the slides you will see a PIN number, so if you could get a little bit prepared when you see the number and just log into Slido so you are ready for when we do the interactive part.
With that, Euro‑IX is a membership association for internet exchange points. We currently have 70 IXP members, while the name suggests euro we do actually have members from outside of Europe and I think this enriches the association and gives us more access that can share the knowledge and them them grow.
We also have patrons and they are typically organisations that work closely with internet exchange points. We have eleven patrons and they are also very valued and definitely part of the membership.
If you are interested in joining as a member, if you are not already a member of Euro‑IX or you are a vendor or data centre or interested in joining as a patron, please speak to me.
So, very quickly, what do we do.
We have two events a year, we just actually had our last event last week in Prague, which was hosted by our members Nick.cz, it was a great turnout, we had 108 attendees and two days of great presentations and networking.
We also have or have started to run workshops throughout the year as well, we had our ‑‑ we had two workshops this year, the first one was a root server workshop which we presented on in the RIPE meeting earlier this year, and last week, while we were in Prague, we had a regulatory workshop which was very excellent because we had a great panel talking about fair share in cybersecurity.
We do a number of services as well, we have reports, we have newsletters, we have mailing lists, these are for the members, we also do benchmarking, again that's for the members. And something new that we have started this year is actually Working Groups so we have a number of Working Groups and I am going to speak a little bit about those later on.
So, as well as the services we do a number of things for the community, and we have tools and the tools include the IXP database. I have spoken about the IXP database in the past, I am not going to really speak about it today but if you are an IXP and you want to participate in the IXP database, please speak to me. If you are using the IXP database for peering information, also, please come and speak to me, if you have any questions about how to access the data or the API.
We have a Peering Toolbox, we also fellowships in mentor IX programmes and this is for IXPs that need support. Again we have got the Working Groups and earlier this year we also released a new film about IXPs so if ‑‑ like for me, my friends and family sometimes really don't understand what I do so I tend to kind of point them to a little video and say watch this and it will give you some idea of what I do and what's involved.
So, please take a look at that and again, if you have any questions, come and see me.
So, now, I am going to move on to our member news updates. So what we do is we invite our members to give us any information that they want to share with the Connect Working Group so I am going to go through that now.
The first update we have is from DE‑CIX. They have started to launch new IXs in the Nordics and you can see the locations there, and they have also recently launched a new IXP in Mexico, South America.
They ‑‑ DE‑CIX is also opening Cloud exchanges in key Cloud hubs offering connectivity, solutions and enable seamless on demand direct access and those are in Amsterdam, Tokyo and Osaka at the moment, which are planned.
A couple of traffic peaks for the DE‑CIX internet exchanges, we have seen peaks in Dallas, New York, Lisbon, Madrid, Marseille, Munich, Frankfurt and Dubai. So congratulations to DE‑CIX on those peaks.
Next up we have BCIX which is the Berlin internet exchange, they have finished doing their VXLAN and EPVN migrations and enabled ND proxy, they have turned up their first 400 gig customer ports and refreshed their brand and logo and that's the new logo there and they have got it in different colours as well so that's a nice little thing there .
Interlan hosted our forum earlier this year in Romania, which was a really great location. They are also see traffic peaks of over 500 gig. And they have 100 gig ports available in both Bucharest and Frankfurt.
Work in progress to implement Nokia routers in the core and convert to EVPN and they are adding support for 400 gig ports so busy times ahead.
AMS‑IX, some of you may have noticed a small little outage on an internet exchange in Amsterdam, this was on 22nd and 23rd November. The problem was well understood and they are following up with the vendors on permanent fixes. And Stavros is going to be giving a lightning talk tomorrow about ‑‑ to give you more information about what exactly happened. So, if you are interested in that, don't miss the lightning talks.
So that's all the news we have from the Euro‑IX members.
Now, I am going to very quickly tell you about some of the work that we ‑‑ one of the projects that we are working on which is the Peering Toolbox.
So, a couple of years ago I had a number of networks and IXPs come to me and say that we are continuously trying to explain what peering is and what transit is and what an AS number is and I mean there's lots and lots of information available on the Internet, some of you will have got a content as well, but the problem is as a beginner where do you start? So, if you are new to the industry it can be quite overwhelming because there is a lot of information. The idea with the Peering Toolbox it's short sections, with clear explanations and we have also tried to structure the way that you can ‑‑ you can access the information, so for example, we have a beginner's section, an intermediary section and advanced section. The beginners' section is complete although we are always going to be adding to it but the first phase of that is complete and was released earlier this year so if ‑‑ please take a look, it's Peering Toolbox.net, and like I said the beginner's section is done and please share that so you know if you are working with any customers or networks that are trying understand what peering and interconnection is, then it's a really, really good tool.
The intermediary section will be worked on next year.
Also, I just want to add in there, there is actually sections in there which include the usage of PeeringDB and the IXP database.
One of the new Working Groups that we have and I thought to give everyone here a little update because you may be interested, is the RFC8950 Working Group. So, we have two Chairs for the Working Group, which is Aleksi from Andrei, they founded the Working Group in the forum we had this year and so far they have had a couple of Zoom meetings, there is a mailing list, the discussions are happening on mattermost and we have GitHub, a GitHub repository where we are also keeping notes and tracking the work that's being done.
So, what's actually been done this this short six months? They have set up a container lab environment for testing, their aim is to test the vendors' BGP implementations and inter operability, they have been done on Arista, Juniper, BIRD and ARouteServer.
Some other points that the Working Group are looking at is forwarding requires special configuration, no IXP peering LAN IP was visible in the trace route. Can RFC8950 peers and traditional peers be safely mixed on the same root server? And again if you are interested in hearing more, come and speak to us. I do have people ‑‑ sorry, there are people here who are part of the company, so if you want to understand a bit more on a technical level what's going on, I'm going to point you to Stavros or Daniel, who is there.
So, yeah, so Daniel and Stavros are involved in the Working Group so if you want to know a bit more about what's being done, feel free to approach them. Thanks.
Okay. And that is all in terms of my presentation.
Oh, any questions? We are going to move on to the interactive section, before you go ‑‑ did everyone get the Slido? The number? If everyone can log in.
SPEAKER: I would like to ask a question first. Could you go back. I am not 100% know that not everyone knows what RFC89350 is all about, could you just tell us?
BIJAL SANGHANI: Maybe you can.
SPEAKER: It is running more or less IPv4 on top of IPv6, everyone knows that we are running out of IPv4 address space and Remco also told about that we already are cutting down the standard assignment for IXPs from the /24 to the /26 and so this Working Group is to explore how we can do it, this RFC explains how to do it that we can do it but I guess there are lots of issues to take care of, and especially I guess most of you are network operators here in the room and we definitely would like to hear from you what an ‑‑ IXPs are here, but I guess most of the IXP probably also are network operators so somewhat related to, at least also if most of them are IXPs please talk also to your customers and make them aware that someone in the next five to ten years we probably each ‑‑ each IXP has to move away from its IPv4 assignment and to only run IPv6 on the peering, thank you.
BIJAL SANGHANI: Thanks for that excellent description, Arnold.
Okay. So I see some of you are already at it. So, I have got about ten minutes, ten minutes ‑‑ well, maybe a little bit more, so I would actually like to make this session a bit interactive. In each of the questions that I have, with polls, there is also, as you can see, something else. There's not a lot of ‑‑ there's not a lot of ‑‑ this is a small group of people so if you are going to answer something else, it would be really nice if you came to the mic and gave a little bit of information about what that something else is.
So, with that, so, yeah, any time you want to come to the mic, please feel free to come to the mic and say what you have to say. The first question I have here is:
What type of entity do you represent?
And this is really just to get an idea of who is in the room, what kind of ‑‑ yeah, who is in the room.
WILL VAN GULIK: In my case it's weird because I am the Chair and I am a network operator and also an IXP and a private person because I am here with my own ASN so I have got these multiple hats and I suspect many people who run community IXP also have side or a real job let's say that's ‑‑ they use to get fed in the middle of the month and end of the month so that's where I suspect you get some of the "something else" there.
BIJAL SANGHANI: Thank you, good point.
BIJAL SANGHANI: Yes, thank you.
Harald: I represent a university.
BIJAL SANGHANI: Next up, we have ‑‑ so obviously this is about peering, it's about interconnection, it's about how ‑‑ we are here to learn, we are here to grow, we want to get better. So, for those of you that are peering, what is missing? What, for you, could make peering better?
I will give a few seconds for people to respond. You can say something else as well, try and get more people on the mic.
So pricing transparency, I don't think that's much of a surprise. Accurate, reliable data, it's good to see that that is high up at the list. Peering policies, maybe, maybe somebody can expand on that. And something else. So Arnold?
Arnold: I am curious why you put in pricing transparency because I always thought that peering is settlement‑free. Settlement‑free for me means I take my cost and you take your cost and why do you want to like how much I am paying for peering, I am not interested how much are you paying for peering but probably you mean something else?
BIJAL SANGHANI: Well so, pricing transparency, so just to be able to know if there are any costs involved with peering, what they are and that's what I meant. Harald.
Arnold: So that would be cost transparency then?
BIJAL SANGHANI: Yes, yes.
Harold: I want to comment on Arnold's questions, I understood this is the pricing of the IXP, the price towards the customers and not between the peers themselves. Is this the case?
BIJAL SANGHANI: Yes, it could be port pricing, it could also be ‑‑ you are right, it's quite generic so maybe I need to hone that in a little bit more.
Accurate reliable data, it's good to see that that is up there. Peering policies, something else, does anyone want to expand on the "something else"? I hope that we are going to get more people than just Will and Arnold coming to the mic.
WILL VAN GULIK: Something else is funny because we need to renew our clothes and we get T‑shirts and the interaction with all these marvellous people.
Arnold: What I also would like to see with peering policies to set up some standards so that peering policies would be available automatically, so currently, I only find peering policy information in PeeringDB, you have an URL and you are categorised in IP with everyone, it's restricted or I do not peer anyhow, but there is a lot of in between and therefore I would like to see something perhaps peering policies scheme at at that so you can automatically evaluate and then check automatically whether the other side would match your peering policy.
BIJAL SANGHANI: Yes, like a standard template that could potentially be used for peers and stuff like that.
Alex: I was working at France IXP, I am an operator so when you are inside an IX you are looking for the technology to get more members but when you are working as CDN network operator you are looking for opportunity to peer so what I am looking here is contact with other peers or introduction, I am going to the middle list, I don't which region and I am looking for the IX who entered this me to other peer so it's peering policy, accurate data, but pricing is my own business, I may go to another provider for that.
BIJAL SANGHANI: Thank you.
Blake Willis: This is pretty cool. In the "something else" I would add two things. First it would be nice to know if IXP is a member based organisation with public pricing that is the same for everybody or if it is purely a commercial organisation and you have to go and negotiate stuff?
The second kind of along those lines would be whether or not the IXP also offered a marketplace‑type portal so that carriers like me that want to connect to these things can make a business case for doing so by saying that we can connect to this and sell stuff?
BIJAL SANGHANI: Yes, good idea, I like both, thank you.
ANDREI ROBACHEVSKY: I don't peer but I voting for peering policies because I agree with Arnold that we need clarity and standard way and standard policies both from IXP and ISP sides and especially security policies such as MANRS and it would also be good if tools like PeeringDB reflects these policies, start working with PeeringDB on labelling some of the sort of, allowing opportunity to say I am a MANRS participant in PeeringDB, I think that would be useful.
BIJAL SANGHANI: Thanks, Andrei. You mentioned MANRS and PeeringDB, I wanted to say in the IXP database we actually do get the API feed from MANRS so if you are curious to see which, obviously you can go to the MANRS website but if you are on the IXP database you can also see which networks and IXPs are MANRS‑compliant.
LEO VEGODA: Speaking as the PeeringDB product manager. So the product committee did discuss whether we should add certifications and things like that and they will definitely look at it again but the decision taken was that adding a certification should only be done in PeeringDB if that would influence whether someone would peer with you so I think this is ‑‑ going back to the question that Arnold asked earlier: If you would only peer with someone if they were certified by MANRS or had some other certification, then please let the product committee know because that will influence their decision, they can review what they have decided in the past. Thank you.
BIJAL SANGHANI: Thanks, Leo.
So, the next question I have is:
What strategies should IXPs ‑‑ and I know we have got a lot of IXPs in this room so this will be really quick and we are running out of time ‑‑ so what strategies should IXPs pursue to respond to industry challenges?
So we have automation, cost reduction, added services. I would be curious to know what services IXPs should be looking into. Getting new customer segments on to the IXP; something else.
Does anyone have anything ‑‑ it's nice to see automation up there at the top, actually.
BLAKE WILIS: Obviously I will say marketplace as one of the added services
BIJAL SANGHANI: Yes, marketplace. Actually, there are a quite a number of IXs that are offering that service now. I guess there's no directory to say these are the IXs that offer that service and maybe that might be helpful.
Arnold: I would like to add to the automation I do not know whether we have seen already a presentation on IX API in the Connect Working Group but most of you probably would have seen it for example, API at NANOG and so on so there is already a piece of software called IX API which just does exactly that, that customers and partners are able to automate a request and everything, why an API towards the IXPs and at least exchanges like AMS‑IX, LINX and DE‑CIX, Netnod are already supporting it.
BIJAL SANGHANI: Thank you.
WILL VAN GULIK: So, I am on the automation with IXP manager and that's what I suggested to some part of the new comers that will try to open an IXP soon in Lee on, that's the way to do it, you will not have to do any workish and everything will be automated and easy and we can generate or filter list correctly and that's where it makes things easy and also then we can have related data and content and we can trust the place where we connect to.
BIJAL SANGHANI: Thank you, IXP manager is a great tools for IXPs, I think there are over 200 IXPs now using it so I think that's a testament to the tool so yeah thank you, INEX, for the work you do on IXP manager.
Okay. Next up:
What is your main challenge when managing peering?
Peers lacking technical knowledge; peers lacking business knowledge, communication with peers about peering, internal lack of understanding of peering ‑‑ within your own association ‑‑ access to accurate data.
And that's all I see there.
There should be a "something else".
STAVROS KONSTANTARAS: From AMS‑IX. I am missing the option "all of the above".
BIJAL SANGHANI: That's the something else. But thank you.
BLAKE WILLIS: Also missing the option "cost".
BIJAL SANGHANI: Thank you. I mean, so here I just want to give a little ‑‑ I will let Arnold talk first.
Arnold: Yeah so heading for the communication with peers about peering, my wish is the same with the peering policy, so currently I guess if an IXP announce a new peer at an exchange the new peer gets 100s of thousands of requests I want to peer with you. There is no currently no schema or standard how to announce your policy, your information, to the other customers and perhaps in the end you would be able to set up a new peering automatically because you'd simply exchange or there would be a platform and you say I want to peer with you and the other side okay, I also want to peer with you and then on both sides the session comes up and you are done.
BIJAL SANGHANI: Thank you.
FLORENCE LAVROFF: Working Group co‑chair. No affiliation. I would like to comment on the cost comment made previously by Blake. I would understand that this could be included in the scalability which now appears like down the list, because a scalability is also, even for most a question of cost.
BIJAL SANGHANI: Yes. That's a very good point. Thank you. And just one thing I want to add when we have communication with peers about peering I just want to mention the Peering Toolbox here again because as I presented earlier this is a good tool to to learn about peering and also something that you can point your peers to, to learn about peering if you are getting questions.
Okay. Which of these will affect your business the most in the next five years:
Regulatory compliance; security and privacy; scalability again; latency and performance; environmental sustainability; edge computing.
Again, sorry, there's no "all of the above" but there is a "something else".
WILL VAN GULIK: We will come close to the end, we need to be fast or we have got roughly, two three minutes, then we need to close the session.
BIJAL SANGHANI: I am almost done. I guess no surprise that regulatory compliance is top of the list there.
WILL VAN GULIK: I was the one who said ‑‑ scalability, I think if we don't have ‑ to peer that would be something to consider, it should be what that we can extend at some point.
BIJAL SANGHANI: Thank you.
I am running out of time so I am going to do this for one minute so if you are typing, be quick:
Name one word that comes to mind when you think about the future of interconnection.
I am going to move on.
So you know, we have had, we talked about this at the beginning of this session and Connect Working Group has been around for ten years and there are other peering events that people go to like the EPF or peering days, certain NOGs, there's all this, different different events of IXP, member meetings, things like this. So, what I'd like to know is, what format would you like to see for peering topics at peering events?
So not necessarily even formats a but topics, would you like to see that you think are interesting and would encourage you or motivate you to attend those events or those sessions a bit more?
So, okay. We have strategy related presentations; panel discussions; tools and automation; training, BCPs, peering personal ‑‑ that has been you love it or you hate it, I was talking to Florence about this earlier and I think there's definitely value in the opportunity for people to come and introduce themselves and let them know where they are from, but also, it's a fine line.
Is there any ‑‑ right, so I am out of time now, I think. Unless there is any last comments or anything at the mic?
Okay. In that case, thank you, everyone, very much, thank you for participating in the session, I will take all this back and my plan is to actually review the answers from the different events that I have carried out this interactive polls and report back at some point, maybe next year. So thank you.
FLORENCE LAVROFF: Thank you, everyone, conscious of time and conscious that we stand between you and your lunch, time to get that session closed.
WILL VAN GULIK: With this, I think we will see you at the next RIPE meeting, wherever ‑‑ Poland, Krakow, something like that.
FLORENCE LAVROFF: Exactly.
WILL VAN GULIK: With this, thank you very much. Don't forget to rate the talk, that's really useful for us and Remco, anything to say?
REMCO VAN MOOK: That was going to be my exact comment, don't forget to rate the talks. And hope to see you all again next time.
WILL VAN GULIK: Thank you, everyone.