We're sorry but this page doesn't work properly without JavaScript enabled. Please enable it to continue.
Feedback

Can we measure the (de)centralisedness of the Internet with RIPE Atlas?

00:00

Formale Metadaten

Titel
Can we measure the (de)centralisedness of the Internet with RIPE Atlas?
Serientitel
Anzahl der Teile
644
Autor
Lizenz
CC-Namensnennung 2.0 Belgien:
Sie dürfen das Werk bzw. den Inhalt zu jedem legalen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen.
Identifikatoren
Herausgeber
Erscheinungsjahr
Sprache

Inhaltliche Metadaten

Fachgebiet
Genre
BildschirmfensterInternetworkingTermFokalpunktClientStrom <Mathematik>ServerEichtheorieRechnernetzInternetworkingRechenschieberGüte der AnpassungRahmenproblemEinfach zusammenhängender RaumKreisflächeBenutzerbeteiligungFormation <Mathematik>KonfigurationsdatenbankProgrammierungSelbst organisierendes SystemNetzadresseExpertensystemSoftwareIn-System-ProgrammierungAutonomes SystemCoxeter-GruppeZentralisatorVideokonferenzBitPunktFacebookGoogolServerEinsClientRichtungSchnittmengeRechter WinkelComputeranimation
RechnernetzBitMatrizenrechnungEinfach zusammenhängender RaumSoftwareInternetworkingComputeranimation
RechnernetzTelekommunikationEinflussgrößeStichprobenumfangPunktNormierter RaumRechenwerkInternetworkingSchätzungPolygonnetzAliasingQuantisierung <Physik>Einfach zusammenhängender RaumQuaderSoftwareEinflussgrößeSchätzfunktionMereologieDatenbankCookie <Internet>PunktWort <Informatik>GrößenordnungTelekommunikationProjektive EbeneZahlenbereichStellenringInformationPolygonnetzTypentheorieURLAdressraumQuick-SortMomentenproblemQuellcodeRoutingOffene MengeInternetworkingSelbst organisierendes SystemNetzadresseVorlesung/KonferenzComputeranimation
Einfacher RingDiagramm
ComputeranimationDiagramm
SoftwarePunktMAPVisualisierungEinsEinflussgrößeService providerHybridrechnerMereologieInternetworkingKreisflächeRoutingGruppenoperationDiagramm
InternetworkingDatenmodellRechnernetzEinflussgrößeDatenpfadÄhnlichkeitsgeometrieCodeWeb logEinflussgrößeSoftwareZentralisatorVisualisierungInformationAnalysisValiditätCodeTVD-VerfahrenEndliche ModelltheorieÄhnlichkeitsgeometrieInstantiierungURLt-TestIn-System-ProgrammierungXML
Computeranimation
Vorlesung/Konferenz
Web ServicesKollaboration <Informatik>Flussdiagramm
Transkript: Englisch(automatisch erzeugt)
Yes, yes, yes, I know, don't move, don't turn anything, or you change, where?
So I will be giving you the time, do you want me to warn you, five minutes before we go?
Five minutes? Okay, yeah, I'll have to be fast anyway.
Good morning everybody, my name is Emil Aben, I will be talking about
measuring the decentraliseness of the internet with Wipe Atlas. Quite a long title, quite a long name for a dev room. I also have a lot of slides, 42 slides actually, but this is because this is actually two presentations. This is going to be my presentation. I also have a colleague here, Vesna, whose talk unfortunately wasn't in the program,
which will be at the end of my slide set. So you'll only get half of the slide set that I'm going to present today. So I'm Emil, I work at the Wipe NCC and we do something called Wipe Atlas. So first a couple of disclaimers, my expertise is in measuring the internet.
I'm not an expert in the decentralised internet, so I'm a bit nervous to be talking in front of a room of experts on that. So I hope I'm getting the framing right. So what I want to focus on is decentralised infrastructure.
So the internet, the infrastructure that's already there. How centralised, decentralised is that and how can we measure that. So current internet is, a lot of it is client server. These servers are centralised, you have the Facebooks, the Googles, that everybody knows. But the original idea of the internet was the end to end principle.
So you connect users directly. And that's something that I would like to measure and see how that actually, what that looks like in countries. Do we see centralisedness or decentralisedness there. So I put down a toy example, two networks there, the circles.
These have, so the organisation I work for is the registry for IP addresses but also for something called autonomous system numbers. So every network on the internet, every ISP has an autonomous system number which allows them to connect to other ones.
So this toy example, you have two of these, say, two ISPs and they connect via something called an IXP. So that's an internet exchange point which basically is a big switch. So say these networks have 50% of the users each.
So, but what actually matters is not that these networks exist but that there's connections between things, connections between users. That's what makes this internet useful. I was thinking a little bit about it and I was like, yeah, these networks actually see 75% of these connections. If you look at the matrix there, and I'll step out so that I'll be private from the camera there.
But if this is all your connections, this is the sources and destinations, then if you, this is actually what AS7 sees.
So that's 75% of these connections and these connections are, I think, the important thing. The other network, AS42, sees the other 75% and the IXP only sees the stuff that is between these networks, so it sees 50%. So number of users is not the number of connections and I sort of wanted to start measuring that type of stuff.
So this betweenness or this IXP is between, it's 50% between the networks. So how do you actually measure that type of stuff? And here, for a moment, I'm assuming that all user-to-user communication channels are equally important.
It's not just rich white guys that are important, it's everybody is important. So we need vantage points in all these user networks. The second one, the second ingredient is estimations of populations in these networks.
So you know you have a network, you have a vantage point from which you can measure this. You need to know how many users these points represent and you need a tool that measures all the things, all the networks, all the infrastructure between these vantage points.
And I put that picture there because this is not very, this is quite messy. This is as messy as baking cookies with my kids. So I want to impress on that this is not exact. This is, I'm trying to do the best with the tools and estimates that we have.
First ingredient is RIPE Atlas. So that's a project I'm involved in that's putting these little TP-Link boxes and other boxes all around the world to actually measure things. I have a couple with me.
So if people are interested in putting them in their networks, they're here for you to take. And what this does is measure the network by community for community. So you can, if you are part of the network, you can also run measurements on it and it runs a couple of measurements by itself.
And it's on Wikipedia, so you can actually find all the useful information there as well. So two more ingredients. There's the user populations of the networks
and I'm using some data from a sister organization, APNIC. They measure roughly how many users there are in networks. I mean this is very rough. So it's order of magnitude measurements. So there's a URL there.
And another ingredient there is something I call IXP Country Jedi. That's a piece of software that we wrote at the RIPE NCC that basically takes these probes and does a mesh between them from all the networks in a given country.
So then with TraceRoute you measure the path in as far as you can measure it. That's where the messiness, one part of the messiness comes from. And then we analyze that. We also add in something called OpenIP Map, which is a database of infrastructure IP addresses
The IP geolocation databases are typically for end users. The stuff in between is not very well represented there. So that's something we are trying to solve. Also a very interesting project, I think. As I said, many caveats. I cannot get into that. I have 15 minutes.
So our picture is that what I'm doing is I create sketches of the local Internet, user-to-user connectivity. So that was a lot of words. So now I'm hoping to get it into pictures.
So this is what Belgium looks like. So what you see here is the other ring represents all the users. So you see Belgium, yes. That's probably not readable. Sorry?
I don't know if this is light. Ah, oh yes, yes. Somewhat readable now. So this is one. So according to the data that I received, 50% of the users are in BelgaCom.
I don't know if that's correct, but that's what this part represents. The next one, Telenet, is this many users. Brutele, this many users. And there's a couple of others. So on users, it's centralized in a sense.
But then we do these trace routes between these probes representing these users. And we find bNix. So this is an Internet exchange point that is quite central. We see Bix, which is a large network here. And we see Level 3, which is a big American transit provider.
And the sizes of these circles represents the between-ness of these. So I also have the data in the back for this. And it's colored by, these are end-user networks. The blues are the transit networks and the oranges are the IXBs.
So Belgium, this is what the US looks like. Here, five minutes, I'll just skip through. You can actually see there AT&T is a user network, but also a transit network. So you can have hybrids. This is what Canada looks like. This is what Korea looks like.
And I find this amazing that this is, the way I know the culture there is people are not very direct, but their networks are very directly connected. So that's, I love that part of this visualization. South Africa loves Internet exchange between the users there.
And this is Mexico. That's Carlos Slim's network. We don't have a probe in there, so we cannot measure it. But there's a lot of American networks between the other ones. So can we measure this? Yes, we can create sketches. Next step, validate model.
I like people to poke holes into this stuff. Is this stuff what I'm doing? Does it make sense? For countries, you can also do this for other decentralized networks. If you have large networks, you can, for instance, add Atlas probes, label them.
We have something called a variation of the ISP country JEDI, which we call the hackerspaces JEDI, where you just take all the probes and hackerspaces and do a similar visualization. There's a URL for that. Or you can use similar methodology and data with the tools at your proposal for your decentralized infrastructure.
So more information. That's the ISP country JEDI that contains all this code. So if you want to see this for yourself or your country, this URL and the execs replace that with the country code. I have promised a colleague that I will write all of this up.
On labs.write.net. So if you're interested, please follow that there. And that's it. This was not just me. This was also Jasper doing these visualizations. That's now for ideas and analysis. We have an awesome student, Petros, who was also working on this.
So this was actually work of that team. So that's it for me. Any questions?