P2P Networking Hub - Guided Tour of Ethereum's Networking Layer

Transcript

[music] Okay, let's get this set up. Uh, so my name is Janice. Uh, I represent several organizing teams uh, of the peer-to-peer networking hub. And we are now going to go through what is actually peer-to-peer networking, what do you expect to see there? and um what the days uh the following days will look like at the peer-to-peer community hub. Um so how do we connect? Do I just press >> slideshow?

>> Yeah. Let's see. Yes. In the meantime, as we set up, I'd like to remind the audience, everyone, we use the Mircat application, um, which you can find via the Q&A QR

codes on the side of the screen if you have any questions to ask the speaker, uh, during the session or at the end. We'll, uh, select them and you can also upvote them if you'd like to see them answered. So, thank you and enjoy the conversation. >> Um, thank you. Thanks for, uh, being patient. So, uh, we're going to talk about the peer-to-peer community hub is going to be talking about what is below the chain. Uh, I'm pretty sure many of you have experienced what it looks like to make a transaction on the Ethereum network. Um, what it looks like to have metrics on uh for onchain um things, but this is not what we're going to be looking at uh at the

peer-to-peer layer. Uh the PTPA layer sits below um the chain and is the layer of the network that gets the transaction from your computer from your phone and propagates it to the network so that um the transa transaction is actually executed. And there how there is a whole bit of uh kind of science and protocol work that goes in there which people very often um don't see. So this is what we're going to be talking about. Uh my name is is Janes Charas. I represent the ProLab team uh which is a team that um is monitoring the peer-to-peer network

and has got uh metrics about several of the uh critical path parts of what's happening as we said below the chain when the uh the transaction and the block is propagated in the network. Um um I represent also the rest of the organizing teams of the peer-to-peer community hub which is uh the Ethereum Foundation uh Hopper NosisVPN and Optimum. Um so lots of interesting conversations to be had but first of all let's see very briefly um who is in the room and perhaps what is uh your um involvement in the peer-to-peer layer. So kind of by show of hands uh who here runs a an

Ethereum node uh validator node or otherwise. Okay, a few hands. Um who has played before with some peer-to-peer networking stack like uh LIP2P, IPFS, Ethereum, of course P2P. Okay, more hands. Uh or you've never thought about how a block is being propagated and you kind of don't care about it. Okay, zero hands. Oh yeah, one hand. Good. Um, so a a brief uh a brief note of how things are are happening. So you might have heard that a new block is um is being

produced every 12 seconds. This 12 seconds is a period that the network needs in order to do a few things. And it starts with the four first four seconds out of these 12 where the validators need to see and vote on the block. Uh during this time there is zero uh central servers. That's why it is a peer-to-peer network which means that every node is equal to others. It needs to receive the block and then propagate it further. Uh there is no data center in the middle at least. Um, of course there are nodes that run in data centers and we kind of uh monitor these and we know like um what percentage of the

network is running on uh data center cloud infrastructure. Uh but otherwise um nodes run for themselves. They have their own operators. um uh they travel like all the blocks, the votes and the blobs are traveling peer-to-peer uh and um yeah have to reach everyone in the network in order for the chain to um uh to to progress. Uh this is different to the client server model that most if not all of the other applications that you have been using are uh are built on. In a client server model, you have um the server that is doing something serving a

website, serving uh an eShop or streaming music or or a movie and your node is sitting somewhere in this venue or at home or on your mobile device while you travel and um content is shipped from the server to your node. Um the peer-to-peer model is a little different. Perhaps you know the representation there is not the best but what it tries to say is that you have different nodes each one of the of which is equal to others and can serve the content um itself. It doesn't need to to have um a central uh presence somewhere

in a cloud environment or data center. Um the node is what is um every node is doing both receiving content and sending content. Um so what are the benefits of doing things in a peer-to-peer way? Uh first of all there is no single point of failure. Um like in this previous slide if that node there goes out if it loses connectivity or there is a power outage then basically it cannot serve you.

We've seen even with big massive data centers uh recently that there are outages although they are so well connected and so well um um they have lots of resources. Um in in a peer-to-peer network if a node fails then there are other nodes to kind of cover up. Uh it can scale with users. Okay this is uh with an asterisk. It cannot scale indefinitely, but the more users you put in there, the more nodes you have, uh the more the replication that can happen depending on how you administer the protocols uh underneath.

It's permissionless in most cases. Um at least, you know, in in Ethereum and other blockchains and uh peer-to-peer networks. So, you don't have to ask anyone to join. You just uh set up your node and and go and participate in the network which is quite different to you know if you want to to do anything basically in other um in other applications and kind of uses the edge as compared to the core of the network because you can have a node at home. Uh you don't need to be located in a very centrally uh overprovisioned data center. you can use your own home connection and uh at least be a homestaker for Ethereum uh but also

other um other peer-to-p peer applications. Uh of course you know there like in saying all of the benefits of peer-to-peer there are some down world downsides there are some challenges as well. Um having a node at home means that if your electricity provider decides that it's a good time to you know do some maintenance work and you are out of electricity then your node goes out which is what we call churn. Um it's also limited in terms of it can be limited in terms of resources.

So uh the network can be as fast as the network's kind of slowest link so to speak uh because you know uh everything needs to propagate to everyone in the network all the blocks and transactions as we said. So if a link is very slow then that slows down you know the propagation and the um pro progression of the chain. Um there is no in most cases and applications there is no authentication. So uh this is a problem because it can like allow some surface to have uh civil attacks. So you can um uh pretend to be someone else and in that case of course security uh kind of suffers and we need

to have the right techniques in place in order to avoid this from happening. um privacy like you can see your own traffic but this like being connected peer-to-peer means that you're also sending traffic to other peers you are the server as well so other peers can see you and the traffic that you're sending not necessarily not always again it depends on the protocol uh setup and protocol configuration but um it has to be done with care because you're more exposed given you play a more active role in the network disappear. Um it it can be unreliable. Uh as we said uh you know a home connection can go down

because electricity went down or your ISP just went down. Uh it's not like massive data centers where you can have multihome things and have several different connections. um nuts. Uh for those of you that have um heard about the term network address translation boxes, which is what um provides quite a bit of security for our personal data and computers at home, but also makes it difficult to um connect to others in the network and have others connect to us. Basically, like have an incoming connection from the outside world into our um home computer at home.

uh which of course makes it difficult for homestakers for example. Um so um peer-to-peer as a networking technology has got a quite a long history. Uh basically I mean it was mentioned in some uh talks yesterday or today. Um the the internet itself was peer-to-peer once. Um then it wasn't anymore. And there is an interesting history about this uh about why this happened. If you're interested, you can come to the hub and we can have a a discussion around it. Um, but basically like from more well-known applications, there was Napster somewhere in the '9s, late 90s, which was file sharing,

several other um versions of it with Bit Torrent being the most uh well-known one. Uh, Kadimlia is a DHT structure that is very widely used and very efficient. it is being used by Ethereum and uh several other networking protocols such as lip also used by Ethereum. Um Bitcoin was yeah we all know what Bitcoin is is a peer-to-peer uh system as well uh built on like um you know computers connecting in a peer-to-peer fashion. Uh then there was IBFS and Ethereum in 2015. Um which many of the audience I guess is familiar with and um you know uh has sparked quite a bit of

what web 3 actually is. Um then okay, Ethereum went on to be um you know the picon chain and then peer dust which I'm pretty sure you've heard about uh very much uh relies on like having a sound and technically correct peer-to-peer network. So uh all that to say that there's quite a bit of history. Um what happens like in the network in a peer-to-peer network which is different to the client server model is that you have to find things which makes it more difficult because you don't just go to one server and ask can I have this web page or something

you have to ask everyone in the network in a sense uh at the same time you can't ask everyone in the network because that's not very efficient so there needs to be a uh sophisticated protocol about how you can find things in the network either other peers or content if uh that's what you're after. Um you can spread you you have to spread new information. So as we said the block the transaction has to spread in the network so that others validate it and and vote on it and um and do all the other things and then you have to fetch data from the network perhaps data you missed. Um and for all of that there are different protocols like um as I said to find peers you use a like a DHT structure a

distributed hasht structure is uh is very efficient if you want to spread information uh pub sub protocols are very efficient uh Ethereum is using the lipidp gossip sub protocol which we're going to cover quite extensively in the hub um and then for fetching you can have um several different protocols like a request response or something that like bit swap which is what IPFS is using. Um on the Ethereum world you have two layers the consensus layer and the execution layer. So every node in the network needs to run um two clients one for each one for the consensus layer one of for the execution layer and basically you have two networks. Uh this sounds a

little um more complicated than it actually is although it adds you know complexity to um natively to to what a node is doing. Uh and there are several different protocols used uh there is dev P2P which was initially there when um like in the beginning of Ethereum um with this discovery version four and five. Uh then at the consensus layer you have again discovery v5 but you also have gossip sub and other li2p based protocols.

Um so several like uh advances in um in ethereum make things even more challenging. Um so bigger blocks add quite a bit of like uh more bytes to be transfer to be um to be transferred by nodes in the network. Uh more blobs like pearas and and all of that. Uh there there was a big discussion about what all of this uh is and if like the nodes especially homestakers are going to be able to um to transfer this data efficiently. uh shorter slots make you know like these 12 seconds that we said in the beginning

and the 4 seconds may tighten things up. So nodes need to be faster and of course uh privacy is a thing that um um everyone wants but u you know uh a good a good discussion item for the hub uh which we're going to have a talk about is are they actually anonymous? Um so the hub is kind of shaped by people who show up meaning that um is not a set um list of talks and and and workshops. We can improvise. We want to be as um um kind of helpful and valuable to the community. So show up and come with questions or things that you want

to investigate. Uh there are there is an interesting uh there is a very um a very nice lineup of speakers already but of course you know addressing community feedback is going to be uh is what hubs are here for. Um so yeah if you want to learn uh about how nodes find find each other how a block propagates how blobs uh and data are being avail being made available in the network uh or simply how a peer-to-peer application is built then um yeah come along and uh we're going to address many of these um also think about what problem you have run to

um in at the peer-to-peer layer if you if you have had one and come along because there are going to be a lot of experts in the room to um try and address your problem or or help you overcome it or not have it the next time around. Um, as I said, there are talks, there are roundts, there are hands-on sessions, uh, but there's also hallway time and and people there to answer questions and, um, um, yeah, make it as valuable as possible for you. Um as I said the organizing partners are Ethereum Foundation, Sigma Prime is um participating, GnosisVPN or Hopper, Optimum, Mega Labs um and ProLab and the MOY protocol. Um generally the hub is

going to be of course staffed from like uh every day. There are four days. Um the first one which is today in the afternoon starting after lunch uh is about foundations and discovery. We're going to have quite a few um dashboards showing up um for uh chain scope from from Shaba u mele discussions then mealabs are going to show their own uh metrics for for the network. So uh this is going to complete the day mostly then scaling and broadcast um what like uh the ch what are the upcoming challenges for Ethereum's peer-to-peer network um

what do bigger blocks mean what quick slots mean um where do we stand with regard to uh postquantum and what does this mean for the network layer uh waggle is a new protocol that uh is being developed um And uh yeah and then what do you send appear when the pieces uh are coded like network coding if you if you've heard anything about that. Um yeah how do we make the thing faster? Uh day three on privacy resilience and open problems. We're going to talk about lipop's gossip hub and improvement proposals there. Uh validators privacy which is um what we um briefly mentioned

in one of the previous slides. We're going to present sonar um a tool that is very useful is developed by our team prolab and is very useful if you want to find out how other nodes see your node from a network perspective like uh if it's visible and everything else. Uh but we're also going to have an open problems round table. Um day four uh it's going to be um um primarily dominated by Hopper and the Gnosis uh VPN. So if you're into VPNs using VPNs or want to learn more about it including having a hands-on workshop um then this is the day to be there. Um yeah and with

that uh I would like to get any question and welcome you to the hub. Thank you. Are we are we displaying the If you have any questions, please use the QR code next to the screens. The Q&A uh code you can see there. And we have several coming in. >> Um there's one. Yeah, there's one the one I just uh selected. How uh you talked about dark matter, how does it affect it? And there was another question which I don't see right now. Um that was about the EF experiments with running nodes over tour. I don't know if it's related. Perhaps you could also share some insights on that.

>> Can can you repeat the the question that he's not here? >> Uh there was a question also about uh running uh Ethereum nodes over tour. There was at connect there was an announcement that EF would start research into that. Um and I was wondering if it's related to this question perhaps as well. Yeah. >> Um yeah. So we haven't seen too many nodes running uh behind door for the Ethereum network. Uh we have seen for other networks. So I think um I'm pretty sure it happens um or some people have tried it but it's not a large population of nodes um so far. Uh so how uh the question is how peer-to-peer network is utilizing that dark matter you mentioned. Can you explain that a little

bit? Um so by dark matter uh there I meant primarily the the edge of the network. So the dark matter could be you know my laptop that is sitting at home or in the office is not supposed to be produ providing and contributing resources to the network. It's not like you know the big Google cloud or AWS data center. uh but if you do things peer-to-peer then inevitably what is happening is that I contribute my resources through my laptop to the network. So that's a metaphor kind of the the dark matter. Um and a peer peer-to-peer network is actually

utilizing that. Um I don't know if that answers the question but happy to um uh discuss more. And then the second question is um Q Bastion. Okay, I don't know that a postquantum secure messaging application. How can we integrate peer-to-peer networking while ensuring privacy offline messaging and reliability? Uh it's it's a tricky set of uh things that you would have to do. I would be more than happy to chat with you if you can find if you can find me at the hub. Um basically the peerto-peer layer, the networking layer is uh composed of a number of protocols. It's not really just one single protocol that is being used. So you have to do the

right configuration in order to ensure that you have you know performance you the performance is good because the chatting application you don't want people to be you know messages to be delayed so to speak. Uh so you want it to be performant and that's a specific set of protocols you have to use. uh privacy you have to do specific things to avoid saying you know that my device have sent this message to this other device. So it has to be kind of masqueraded in a specific way and there are techniques about these. Um the most tricky part is actually how do you do the configuration?

Um offline messaging. Yeah, again that is even more tricky. But there are you know Bluetooth like techniques where you can um you can propagate messages of course locally, right? So if we're if there is no outside connection to the outside wall, then our devices might be able to communicate in the venue when we're close to each other, but we won't be able to communicate outside of these. Uh I guess that's what you mean by offline messaging. Um yeah, and reliability again is a is a tough thing for peer-to-peer networks because nodes might go offline all the time. So you need to have again the right configuration to have backup. um uh

nodes to propagate messages. Um what are the most imminent scaling wins you're expecting uh you're excited about for Ethereum? So I think the uh the scaling needs so bigger blocks and tighter slots like shorter slots is what is going to kind of put more stress in the network and we'll have to see how we can um how we can accommodate that in the at the peer-to-peer layer using gossip sub or like enhanced versions of of gossip sub. Uh that's one thing that I'm very interested in as well as making some of these uh adding anonymity. Uh so

there are techniques being um talked about and trying like being uh developed so that there is more anonymity at that layer. Uh that's also something that I would like to see happening. >> Thank you. There are no further questions. Um, unless there are some [music]