camp2023-57164-eng-SCION_for_a_greener_Internet_opus.vtt

WEBVTT

00:00:00.000 --> 00:00:10.000
 [MUSIC]

00:00:10.000 --> 00:00:20.000
 [MUSIC]

00:00:20.000 --> 00:00:36.400
 Hello everybody, welcome to the last talk in this block.

00:00:36.400 --> 00:00:41.400
 Who of you sent your stuff to the camp with BGP?

00:00:41.400 --> 00:00:43.600
 Any user of BGP?

00:00:43.600 --> 00:00:45.800
 I used it, it works great.

00:00:45.800 --> 00:00:51.480
 This is the routing protocol at camp for parcels.

00:00:51.480 --> 00:00:53.120
 But there's another BGP,

00:00:53.120 --> 00:00:56.640
 which is the routing protocol on the Internet for packets.

00:00:56.640 --> 00:01:00.480
 It's what has enabled the Internet for decades.

00:01:00.480 --> 00:01:07.880
 Mentor is going to tell us about what might replace BGP on the Internet,

00:01:07.880 --> 00:01:13.520
 and how Scion might actually help to make the Internet greener.

00:01:13.520 --> 00:01:16.520
 Mentor, stage is all yours.

00:01:16.520 --> 00:01:18.080
 Thank you.

00:01:18.080 --> 00:01:20.920
 [APPLAUSE]

00:01:20.920 --> 00:01:23.720
 In the next 20 minutes or so,

00:01:23.720 --> 00:01:28.840
 I will talk about how the Internet actually works currently,

00:01:28.840 --> 00:01:34.120
 and what the Scion project in particular wants to change about it,

00:01:34.120 --> 00:01:36.120
 and how that works.

00:01:36.120 --> 00:01:39.320
 I will gloss over a lot of the technical detail, unfortunately,

00:01:39.320 --> 00:01:41.320
 because I have very little time.

00:01:41.320 --> 00:01:43.320
 And I will try to be inclusive in this talk,

00:01:43.320 --> 00:01:52.320
 which is why I will use metaphors that might stretch reality a bit.

00:01:52.320 --> 00:01:57.120
 I simply want to give you, those of you who have no idea actually

00:01:57.120 --> 00:02:01.720
 how the Internet works, some idea of how it does.

00:02:01.720 --> 00:02:05.720
 And for this reason, please bear with me if you have an idea

00:02:05.720 --> 00:02:11.720
 what I'm talking about, and the analogy doesn't quite fit.

00:02:11.720 --> 00:02:13.520
 All right.

00:02:13.520 --> 00:02:16.920
 So this is the outline of my talk.

00:02:16.920 --> 00:02:22.920
 First, we need to talk about how the Internet currently works.

00:02:22.920 --> 00:02:25.920
 The status quo is that the Internet pretty much works

00:02:25.920 --> 00:02:30.920
 like a run-of-the-mill postal system.

00:02:30.920 --> 00:02:41.120
 Say, I want to send this plush owl to the Chaos Computer Club in Berlin.

00:02:41.120 --> 00:02:42.120
 So what do I do?

00:02:42.120 --> 00:02:46.920
 I simply put it in a box and add a sticker to that box

00:02:46.920 --> 00:02:51.320
 with a destination where I write the postal address

00:02:51.320 --> 00:02:53.320
 of the Chaos Computer Club Berlin,

00:02:53.320 --> 00:02:59.720
 give it to the nice folks at the BGP,

00:02:59.720 --> 00:03:03.720
 I think they're called the Camp Parcel Service or something,

00:03:03.720 --> 00:03:10.320
 and then they worry about the rest.

00:03:10.320 --> 00:03:14.520
 I don't actually know the route which this packet will take.

00:03:14.520 --> 00:03:17.720
 I only know that at some point in the future,

00:03:17.720 --> 00:03:21.920
 I think a few days or so, if everything works as expected,

00:03:21.920 --> 00:03:30.320
 my owl will arrive safely at Chaos Computer Club Berlin.

00:03:30.320 --> 00:03:36.320
 So I don't know whether this packet will be forwarded

00:03:36.320 --> 00:03:38.920
 via ZNNIC or via Fürstenberg.

00:03:38.920 --> 00:03:42.520
 Maybe it can also take a detour through a logistics center

00:03:42.520 --> 00:03:43.920
 in Frankfurt Oder.

00:03:43.920 --> 00:03:49.320
 I simply don't know.

00:03:49.320 --> 00:03:53.120
 Interesting.

00:03:53.120 --> 00:03:57.920
 I think I need to restart my slides

00:03:57.920 --> 00:04:01.120
 because I just rebuilt them.

00:04:01.120 --> 00:04:18.320
 A short timer, please.

00:04:18.320 --> 00:04:38.520
 Here we go.

00:04:38.520 --> 00:04:42.720
 So this is the perspective of the user that sends a packet

00:04:42.720 --> 00:04:44.520
 over the current Internet.

00:04:44.520 --> 00:04:50.120
 They simply know where to put the packet, the next postal office,

00:04:50.120 --> 00:04:52.920
 and some magic happens in between

00:04:52.920 --> 00:04:58.320
 and the packet arrives at its destination at some point.

00:04:58.320 --> 00:05:01.720
 How does the postal system do this?

00:05:01.720 --> 00:05:04.120
 Well, in very simple terms,

00:05:04.120 --> 00:05:08.320
 and this is unfortunately not exactly how it works,

00:05:08.320 --> 00:05:11.120
 unfortunately, let's assume that there's a post office

00:05:11.120 --> 00:05:14.920
 in every town and each has a few courier connections

00:05:14.920 --> 00:05:17.920
 to the surrounding post office in neighboring towns.

00:05:17.920 --> 00:05:21.520
 And each post office also keeps a huge list of all the towns

00:05:21.520 --> 00:05:28.120
 in the country where it can look up into which neighboring post office

00:05:28.120 --> 00:05:31.120
 it should send a packet in order for the packet to move closer

00:05:31.120 --> 00:05:33.720
 to its destination.

00:05:33.720 --> 00:05:42.920
 So we send a parcel, this parcel will get forwarded intuitively

00:05:42.920 --> 00:05:45.120
 step by step close to its destination

00:05:45.120 --> 00:05:48.720
 until at some point it arrives in its destination city

00:05:48.720 --> 00:05:53.720
 where the postman knows exactly how to find the street address

00:05:53.720 --> 00:05:56.920
 for this packet.

00:05:56.920 --> 00:05:59.720
 So this distributed mechanism, this is pretty similar

00:05:59.720 --> 00:06:05.320
 to how postal systems used to work in the past for several hundreds of years

00:06:05.320 --> 00:06:06.720
 and works very well.

00:06:06.720 --> 00:06:10.920
 A given post office only needs to know in which direction

00:06:10.920 --> 00:06:13.920
 it needs to forward this packet and at some point the packet

00:06:13.920 --> 00:06:18.920
 will arrive actually in the city where it is intended to arrive in.

00:06:18.920 --> 00:06:21.720
 So in there it will be given to a postman or woman or person

00:06:21.720 --> 00:06:31.720
 who will know how to deliver this packet.

00:06:31.720 --> 00:06:36.720
 So resolving this analogy, the post offices are the internet core routers,

00:06:36.720 --> 00:06:38.520
 the post offices.

00:06:38.520 --> 00:06:45.920
 The local post office is your ISP which is where you will send your letters

00:06:45.920 --> 00:06:50.520
 or give up your letters and receive new letters.

00:06:50.520 --> 00:06:54.520
 And all post offices along the way are their own networks,

00:06:54.520 --> 00:06:59.120
 their own autonomous systems, you have probably heard this term already.

00:06:59.120 --> 00:07:04.920
 And all the neighboring post offices of a given post office are called its peers.

00:07:04.920 --> 00:07:09.120
 And this huge list that each post office needs to maintain

00:07:09.120 --> 00:07:14.520
 in order to know along which next hop a given packet needs to be sent

00:07:14.520 --> 00:07:16.920
 is a routing table.

00:07:16.920 --> 00:07:21.720
 And the BGP protocol, not those guys over there, but the BGP protocol

00:07:21.720 --> 00:07:29.520
 is used to keep the routing table up to date.

00:07:29.520 --> 00:07:31.520
 So what is wrong with this current internet?

00:07:31.520 --> 00:07:36.920
 Well, depending on your perspective, not that much or maybe quite a lot.

00:07:36.920 --> 00:07:41.720
 So the internet doesn't have the best reliability.

00:07:41.720 --> 00:07:44.120
 So it has like three nines.

00:07:44.120 --> 00:07:49.520
 It has a 99.9% reliability which sounds a lot but actually isn't that much.

00:07:49.520 --> 00:07:56.120
 The good news about the internet, however, is that failures often last a few seconds

00:07:56.120 --> 00:07:59.920
 which all of you have probably noticed that the website doesn't load,

00:07:59.920 --> 00:08:01.720
 you hit F5 and it does load.

00:08:01.720 --> 00:08:04.720
 And you don't know where the problem was, but it's pretty likely that the problem

00:08:04.720 --> 00:08:09.920
 was with this BGP mechanism where a failure wasn't needed to be resolved

00:08:09.920 --> 00:08:12.520
 in order for routing to work again.

00:08:12.520 --> 00:08:16.120
 There are other problems like security problems because BGP as a protocol

00:08:16.120 --> 00:08:20.920
 was designed at a time where there was an implicit assumption

00:08:20.920 --> 00:08:25.320
 that every participant on the network can be trusted.

00:08:25.320 --> 00:08:31.720
 Actually in the BGP RFCs, in the first ones, the words trust or security

00:08:31.720 --> 00:08:36.920
 or malicious or attack or something, you can grab for this and zero hits every time.

00:08:36.920 --> 00:08:43.920
 The BGP was designed with a completely different internet in mind than we have today.

00:08:43.920 --> 00:08:47.720
 Then there are some broken economic incentives.

00:08:47.720 --> 00:08:54.320
 So as we all know, currently legislation, politics is lagging behind

00:08:54.320 --> 00:09:03.920
 and they don't really enforce a proper carbon tax on carbon emissions

00:09:03.920 --> 00:09:10.720
 which means that it's still the case today that you can buy electricity

00:09:10.720 --> 00:09:14.920
 from non-renewable energy at lower prices than electricity

00:09:14.920 --> 00:09:17.720
 that was created using renewable energy.

00:09:17.720 --> 00:09:23.120
 And obviously market forces force network operators which have running costs

00:09:23.120 --> 00:09:30.720
 and need this energy, force them to basically just buy the cheapest one,

00:09:30.720 --> 00:09:33.120
 buy the discount of electricity.

00:09:33.120 --> 00:09:36.920
 And this is obviously a problem, this is one of the points where the internet

00:09:36.920 --> 00:09:40.120
 is currently not as green as it could be.

00:09:40.120 --> 00:09:43.920
 But it's also hard for users to influence, like you don't really have a choice

00:09:43.920 --> 00:09:50.320
 when some network in between decides to buy electricity from coal plants.

00:09:50.320 --> 00:09:54.120
 You're basically stuck with this.

00:09:54.120 --> 00:09:56.720
 Now, what is different with Scion?

00:09:56.720 --> 00:10:01.320
 A lot of things they are trying to address, a lot of these problems,

00:10:01.320 --> 00:10:07.520
 for instance also addressing security aspects, but I won't go into detail here.

00:10:07.520 --> 00:10:12.320
 The key point where Scion differs from the BGP-based internet is that

00:10:12.320 --> 00:10:18.920
 with Scion the senders get to decide along which paths the data gets forwarded.

00:10:18.920 --> 00:10:24.320
 So again in our postal analogy, what we do instead of writing

00:10:24.320 --> 00:10:28.320
 a regular street address there, we write steps there.

00:10:28.320 --> 00:10:34.920
 We say, okay, the first step is the camp parcel service, the packet needs to go there,

00:10:34.920 --> 00:10:37.320
 before the camp parcel service forwards it to Mildendorf,

00:10:37.320 --> 00:10:40.920
 before Mildendorf forwards it to St. Annick and so on.

00:10:40.920 --> 00:10:45.720
 And at some point it arrives at a point where a post person, a postman or post woman

00:10:45.720 --> 00:10:49.720
 can simply read the street address and exactly knows where it is.

00:10:49.720 --> 00:10:55.120
 So this is the postal analogy.

00:10:55.120 --> 00:10:57.520
 How does this help?

00:10:57.520 --> 00:11:05.720
 How, translating this to data packets, what do we gain?

00:11:05.720 --> 00:11:13.120
 We gain as users the ability to boycott networks that we know operate

00:11:13.120 --> 00:11:15.920
 with a dubious sustainability record.

00:11:15.920 --> 00:11:22.120
 And we can also support networks that we know operate sustainably,

00:11:22.120 --> 00:11:25.720
 that have basically given assurances to other parties

00:11:25.720 --> 00:11:30.920
 that they are only buying energy from renewable sources.

00:11:30.920 --> 00:11:35.720
 And in our example, when we want to send this owl to Berlin,

00:11:35.720 --> 00:11:41.920
 and we know that the operator of the connection between Ziedernick and Oranienburg,

00:11:41.920 --> 00:11:46.520
 which is on the way to Berlin, uses a truck,

00:11:46.520 --> 00:11:52.920
 and we also know that the operator that sends traffic or packets or parcels

00:11:52.920 --> 00:11:55.520
 between Fürstenbeck and Oranienburg uses rail,

00:11:55.520 --> 00:12:01.720
 then I can, as a user decide, please route my packets along Fürstenbeck,

00:12:01.720 --> 00:12:04.720
 even if it takes a bit longer.

00:12:04.720 --> 00:12:11.320
 But I know that the carbon footprint of this transaction will be smaller.

00:12:11.320 --> 00:12:15.520
 There's an additional feature, of course, that I should mention.

00:12:15.520 --> 00:12:20.520
 This allows all sorts of things to do with this capability.

00:12:20.520 --> 00:12:24.120
 For instance, for activists it's also interesting maybe to avoid

00:12:24.120 --> 00:12:29.320
 routing via networks that are under dubious jurisdictions, etc.

00:12:29.320 --> 00:12:35.720
 So you can optimize for more things.

00:12:35.720 --> 00:12:40.920
 It's also interesting that CYON has been designed in a way

00:12:40.920 --> 00:12:46.320
 that allows its core routers to operate more efficiently.

00:12:46.320 --> 00:12:50.320
 In the post office analogy, it doesn't really keep these long lists

00:12:50.320 --> 00:12:54.120
 with which it needs to decide where to send on which packet,

00:12:54.120 --> 00:12:59.320
 it doesn't need to keep this list of all the towns in the country.

00:12:59.320 --> 00:13:03.920
 It simply needs to follow the instructions that are written on the parcel.

00:13:03.920 --> 00:13:10.120
 This comes at the expense, however, of 5 to 20% of bandwidth overhead,

00:13:10.120 --> 00:13:11.720
 which sounds like a lot.

00:13:11.720 --> 00:13:15.920
 It's not strictly an issue in wired communication,

00:13:15.920 --> 00:13:24.920
 because increasing the speed doesn't really scale the electricity need of this.

00:13:24.920 --> 00:13:28.520
 If you just put a new laser with a different wavelength on the same fiber,

00:13:28.520 --> 00:13:33.720
 the laser doesn't really need more energy in order to transport more data.

00:13:33.720 --> 00:13:38.520
 But to basically visualize this overhead that this introduces,

00:13:38.520 --> 00:13:42.720
 imagine putting this owl into a parcel where you write,

00:13:42.720 --> 00:13:46.720
 I don't know, the BGP guys over there onto it,

00:13:46.720 --> 00:13:49.720
 and then put it into another parcel and write the next address.

00:13:49.720 --> 00:13:50.720
 Actually, it's the other way around.

00:13:50.720 --> 00:13:52.720
 First, you would write the address of Berlin,

00:13:52.720 --> 00:13:55.120
 put it into the next parcel, which you close,

00:13:55.120 --> 00:13:58.720
 and then you write Oranienburg there, and then put it in a parcel,

00:13:58.720 --> 00:14:03.520
 and then you write, I don't know, what did I say?

00:14:03.520 --> 00:14:06.320
 Christenberg on it, put it in a parcel,

00:14:06.320 --> 00:14:10.320
 and then on the last parcel you write BGP,

00:14:10.320 --> 00:14:14.120
 and then they unwrap it, and so on.

00:14:14.120 --> 00:14:15.720
 This obviously includes some overhead,

00:14:15.720 --> 00:14:21.120
 and this is a visualization of what kind of overhead it entails.

00:14:21.120 --> 00:14:27.920
 The technical details are a bit different, but this is how it works.

00:14:27.920 --> 00:14:30.320
 Also, there's something that we should mention,

00:14:30.320 --> 00:14:33.920
 is there's an embodied CO2 footprint,

00:14:33.920 --> 00:14:39.920
 because SIRON router hardware is different from hardware

00:14:39.920 --> 00:14:43.920
 that needs to run the regular internet.

00:14:43.920 --> 00:14:49.120
 And obviously, rolling out SIRON globally would mean

00:14:49.120 --> 00:14:54.520
 investing this, embodying this CO2 footprint of the data

00:14:54.520 --> 00:14:58.520
 and constructing all this hardware.

00:14:58.520 --> 00:15:01.520
 Actually, numbers, how bad this will be, I'm missing,

00:15:01.520 --> 00:15:03.520
 I don't know about this.

00:15:03.520 --> 00:15:07.520
 I think a rough ballpark is that a SIRON router will be about,

00:15:07.520 --> 00:15:11.520
 somebody about as much energy as an equivalent IP internet router.

00:15:11.520 --> 00:15:16.520
 So maybe this sort of evens out,

00:15:16.520 --> 00:15:20.520
 but you could feel free to research this yourself.

00:15:20.520 --> 00:15:24.520
 A third point which is actively being discussed in the SIRON project

00:15:24.520 --> 00:15:30.520
 is to attach carbon footprint of each individual hop

00:15:30.520 --> 00:15:34.520
 as metadata to the routing information

00:15:34.520 --> 00:15:38.520
 that then allows the user to construct paths

00:15:38.520 --> 00:15:41.520
 that use the minimal amount of CO2,

00:15:41.520 --> 00:15:49.520
 and then let the user preferences cause pressure on network operators

00:15:49.520 --> 00:15:54.520
 to use renewable energy or to prefer the use of renewable energy.

00:15:54.520 --> 00:15:59.520
 Another aspect of this would be that if they are being honest about

00:15:59.520 --> 00:16:07.520
 their current CO2 emissions used for routing their packets,

00:16:07.520 --> 00:16:10.520
 and this changes dynamically basically,

00:16:10.520 --> 00:16:12.520
 for example based on available wind,

00:16:12.520 --> 00:16:15.520
 then using this information could cause,

00:16:15.520 --> 00:16:18.520
 could help us route our packets around,

00:16:18.520 --> 00:16:25.520
 I don't know, how is it called, if the wind is gone,

00:16:25.520 --> 00:16:33.520
 around calm weather patches where there is little wind,

00:16:33.520 --> 00:16:36.520
 too little wind power for the wind turbines,

00:16:36.520 --> 00:16:42.520
 and so we simply avoid these regions and send our packets by different routes.

00:16:42.520 --> 00:16:47.520
 All in all this could sort of create a virtual circle

00:16:47.520 --> 00:16:53.520
 where the network operators try to compete with each other

00:16:53.520 --> 00:16:57.520
 and try to be more green than their competitors,

00:16:57.520 --> 00:17:01.520
 basically reducing carbon emissions overall.

00:17:01.520 --> 00:17:06.520
 Will this be enough? Will this really help us fighting climate change?

00:17:06.520 --> 00:17:15.520
 Well, so the global internet energy consumption in 2020 was 207 terawatt hours,

00:17:15.520 --> 00:17:22.520
 which is sort of creates to a momentary power consumption of about 31 gigawatts,

00:17:22.520 --> 00:17:27.520
 which is double that of the biggest hydroelectric plant in the world.

00:17:27.520 --> 00:17:34.520
 The Three Gorges Dam in China produces about half this amount of power

00:17:34.520 --> 00:17:39.520
 if it's running at full capacity,

00:17:39.520 --> 00:17:45.520
 and this is excluding the power consumption of data centers,

00:17:45.520 --> 00:17:49.520
 so this is only network equipment, it's a rough ballpark figure.

00:17:49.520 --> 00:17:53.520
 And this number is expected to grow to about,

00:17:53.520 --> 00:17:56.520
 so right now it's 1% of all the global electricity,

00:17:56.520 --> 00:18:01.520
 and this number is expected to grow to 10% by 2030.

00:18:01.520 --> 00:18:07.520
 So the bottom line is we don't really know how much Scion will help.

00:18:07.520 --> 00:18:11.520
 We know that taking all the efficiency gains of routers into account

00:18:11.520 --> 00:18:18.520
 and subtracting the bandwidth overhead that it has,

00:18:18.520 --> 00:18:27.520
 we end up at a number of around 10% more efficient,

00:18:27.520 --> 00:18:31.520
 10% less electricity use, let's say.

00:18:31.520 --> 00:18:37.520
 Of course, we didn't really take into account the embodied carbon costs,

00:18:37.520 --> 00:18:42.520
 and we also didn't really take into account the currently still elevated overhead,

00:18:42.520 --> 00:18:47.520
 energy overhead that is required for anti-virus to actually talk this protocol.

00:18:47.520 --> 00:18:51.520
 I mean, you can help.

00:18:51.520 --> 00:18:55.520
 So it probably wouldn't be enough to save the planet by itself, right?

00:18:55.520 --> 00:19:00.520
 But it's one step along the way,

00:19:00.520 --> 00:19:07.520
 and we can't really afford to exclude things that don't promise that much reward,

00:19:07.520 --> 00:19:13.520
 but we can probably use this as an argument to still push towards a technology

00:19:13.520 --> 00:19:17.520
 that changes the way that the Internet is currently used

00:19:17.520 --> 00:19:22.520
 and at the same time empowers users in a way that allows them,

00:19:22.520 --> 00:19:27.520
 allows additional forces to push towards carbon neutrality.

00:19:27.520 --> 00:19:32.520
 In this talk, I have left out a bunch of stuff which we could talk about offline afterwards.

00:19:32.520 --> 00:19:39.520
 Of course, there's a lot of technical detail that I glossed over pretty much.

00:19:39.520 --> 00:19:44.520
 There are questions about how a SCI-internet would be governed.

00:19:44.520 --> 00:19:48.520
 There are also different incentives for different Internet stakeholders.

00:19:48.520 --> 00:19:53.520
 Your ISP might not be that happy if you always decide to send your traffic a certain way,

00:19:53.520 --> 00:19:57.520
 because right now it's them who decide how to forward it.

00:19:57.520 --> 00:20:00.520
 Anyway, how to get involved?

00:20:00.520 --> 00:20:07.520
 So the SCI-in project itself has been under development at ETH Zurich since 2009.

00:20:07.520 --> 00:20:11.520
 It's open source and standardization is underway at ITF also.

00:20:11.520 --> 00:20:13.520
 It's deployed in the wild.

00:20:13.520 --> 00:20:18.520
 The Swiss Secure Finance Network will migrate exclusively to SCI-ins,

00:20:18.520 --> 00:20:26.520
 so all the Swiss banks will stop using the old system and exclusively migrate to this system by 2024.

00:20:26.520 --> 00:20:29.520
 There are additional networks.

00:20:29.520 --> 00:20:36.520
 So one is a production network that is run by a commercial company that is selling this hardware also.

00:20:36.520 --> 00:20:41.520
 And an education and research network that you guys can join.

00:20:41.520 --> 00:20:47.520
 And also there's an experimental playground where you can play around with this technology,

00:20:47.520 --> 00:20:55.520
 but it still uses the regular BGP-based Internet underneath and works via VPN links.

00:20:55.520 --> 00:21:01.520
 You can play around with this under these URLs and get involved.

00:21:01.520 --> 00:21:08.520
 It's probably best if you join this Matrix channel where SCI-in development is discussed.

00:21:08.520 --> 00:21:12.520
 And I think with that I'm about done.

00:21:12.520 --> 00:21:18.520
 You can also reach out to me at this email or this masternode address.

00:21:18.520 --> 00:21:22.520
 [Applause]

00:21:23.520 --> 00:21:27.520
 [Music]