Incentivize what matters by Robert Drost

okay let's give them another round of applause to call down local here to the ethereum if Denver and ethereum ecosystem so thank you everyone for coming we're going to jump into our next session again my name is Jorge Cortes I'm at MC here for eat Denver and I'm representing and coming visiting from Miami Florida one more year excited to be here sharing with all the community I want to give a big shout out thanks to all the volunteers to all the people that's making this happen from volunteers to people on the staff thank you everybody here in Denver for making us feel like home um so next up uh we're going to talk about incentivize what matters uh and what incentivized apis will build the decentralized public good so for that we have Rover Frost here with us on stage welcome Romulus welcome Robert with a round of applause please thank you I'm miked up so I'm good [Applause] hey there everybody so happy to be here thank you everybody for coming by if I'm correct I think vitalik's talking tonight um I believe that's correct very psyched looking forward to that so um I'm going to start first with a question how many people here use the public cloud I'm looking to see how many people are listening because everybody's hand should be up okay I'm talking about something which is sort of a fusion of things that are very interesting to me and should be interesting to the rest of us obviously you know we love blockchain love defy love nfts love all of that but it's also good to look at how blockchains can be utilized in places where there's new money coming into the system as well and public cloud is a space where a lot of money is spent and so I want to break down what does it mean for us to actually make blockchains much more relevant to the public cloud so to start off let me just say I like to say what I'm going to cover first and then go through it and then remind us at the end we're going to do three themes I'm going to try to move quickly because I want to make sure I get through to the end and the nice conclusions the centralizing force behind today's public clouds then talk about what blockchain tech could actually allow us to have a decentralized public cloud and you know it's not a small problem uh public cloud is huge a lot of transactions it's at a scale that is very hard for us to think about when we compare it to what blockchain tech is capable of so it requires some careful attention and then just very briefly today's status and what's next and everybody I've talked about before the talk who's in public cloud is very excited about both the status and next and find some of this stuff I'm talking about pretty surprising so hopefully it's enjoyable to everybody so one level set I wanted to start off from in the very beginning because a lot of times when people hear decentralized public Cloud there's a lot of kind of marketing and people get excited and say oh this new thing is the AWS of web3 the AWS the blockchain and you know I want to talk about the real scale of what clouds are clouds are not just RPC Services into the blockchain layer there's a very rich stack that was established over time of lower layer infrastructure which are the kind of compute storage networking a bunch of platform as a service so load balancers databases things that allow software to interconnect and finally top level applications so things that you actually use as a user or you may use in businesses like salesforce.com that kind of stuff um so in order for us to make this sort of theoretically correct I want to remind people that you know our chief philosopher of Elise ethereum and maybe we'll say one of the top ones of all of blockchains metallic a while back talked about three decentralization axes and it's very important to be kind of precise when we talk about decentralization because decentralization's a little different on different axes I'm going to do two things to his his diagram which is very good but for the sake of this political decentralization is a little bit confusing it sounds like we're talking about voting we're not we're talking about financial financial centralization and decentralization who controls the money who controls the access to make money who controls the ability to get value from the money that's made and then the second is logical decentralization is not so relevant because actually both types are fine depending on the kind of software or thing you're building sometimes you want to have Global synchronization sometimes it's okay to be local so we get this simplified view okay we're going to come back this in a little bit and talk about where does things like AWS and Azure and Google Cloud which we all use fit on this compared to where we want the decentralized public Cloud to be so let's break down because a lot of people think about the public cloud and I've heard this very often they feel public Cloud it's decentralized it's distributed I've got servers all over the world I can set them up wherever I want you know if something breaks down here it's got lots of redundancy I can do things over there it feels fine and so that's completely correct we actually can say for sure that the architecture of public clouds are decentralized and they're done very well and they perform extremely well they're very high bar for us to hit in terms of the decentralized public Cloud but now let's go through these other three things that are all Financial metrics let's look at who controls the Billing System of the public Cloud you may forget but when you originally set up your AWS account it was very convenient you put in a credit card and after that you don't have to think about it again I mean you get a bill or you give it to your CFO or CIO whatever but it's it's a centralized thing so behind the scenes every time you're using the infrastructure on public Cloud it's being metered you're doing your API calls you're doing some datas moving through the network whatever it's being checked and counted by those public Cloud providers this is not a small task and it's a huge centralizing thing that has to be done by them and completely controlled by the public clouds second is pricing provider access very opaque I mean obviously it goes down a little bit but you can kind of look at it from the beginning of the public Cloud to now there's been sort of a splitting the difference the actual costs go down but only the cost savings that we see is a fraction of that and I'll talk about that when we talk about uh good and bad things and then the value earning access if you are a shareholder in the company then you get the value out of the public Cloud because it's actually going to increase your stock price if you are somebody who's actually bringing Revenue to a cloud you're going to end up splitting with the cloud and you're not going to benefit from that so there is this sort of middleman which we know very much is always at risk of run seeking so now going back to vitalik simplified diagram you know we can show the two things the reason why public Cloud seem to be decentralized is because they are architecturally but we our ideal decentralized public cloud is going to fit on this right side and just very briefly you know public Cloud Market is extremely dominated by a couple of key players we all use them the scale of Revenue in the space is a little bit surprising if you've never looked at the financials it's going up 20 30 a year it's now crossed 300 billion across the space and it's going up so from you know not in economics one-on-one point of view but from an econ one point of view so not just Supply demand but demand the cloud is a great place to actually bring Revenue into the blockchain space and to settle in blockchain so this should be very interesting to all of us the other thing I was going to say is in terms of all of us or most of us being open source software developers there's a lot of kind of capture in this space when software developers go out to the public Cloud they're in an asymmetric relationship with the public Cloud providers if they charge through the public Cloud there's a cut taken just like the Apple Store it's been very successful everybody has to pay it and then there's this problem of Open Source software undercutting and so I just show a picture here that you know a long time back open source software which we all would love to be able to do and you know have a lot of control over the kind of projects we work on it has an open core model where you try to have a lot of the software be developed fully in the open lots of Community Based and everybody sort of says yeah there's features that you add that are valuable and companies pay for it and that's the kind of added top layer around it that's the original Vision that started working well in the 2000s then public Cloud came along that ended up getting split and the cloud vendors actually take open source and do their own in-house version of that open source but they're not the original project and you know somebody could tell me if they know other examples but in general like mongodb is a good example very public the open source projects just say you know it's actually a negative effect on the open source space foreign for devs and users because maybe we don't care about the open source companies but devs and users that are using the uh public Cloud I just call it sub-optimal pricing and you know I don't want to come across from this talk that public clouds or evil and other things in fact I think public clouds can definitely participate in bolster up the initial performance and capacity of our decentralized protocols they can make money from it as well but we have to be conscious about what happens when somebody controls all of the resources they have both sides of the marketplace people who are consuming the infrastructure and the software and then the people who are building that software and you get sub-optimal pricing and there's a lot of them that you can kind of look in but like one of the ones I like the best uh is you know cloudflare which is very strongly kind of going after different models and they're actually getting a lot into the decentralized space who here is use ipfs through cloudflare give them a shot out they have a really nice reliable uncensored ipfs which I love that they've stuck with that idea and they support a lot of companies and a lot of independent people with DDOS protection so a great place so if you look on the right side they did a conservative calculation because they didn't want to like be overly extreme but you know when you're paying for your egress fees out of a public cloud two three four cents a gigabyte to get data out in the US and other places you're paying 80 times the cost of providing that and that's conservative okay it didn't start out that way it actually started out more like this three to 4X kind of number that you see in South Korea but that's sort of the progress when you have a monopoly position you start off with a relatively low you know tax on top of it and then you increase it over time that's what rent seeking Behavior looks like so just calling that out I mean public clouds are very good and I'm not saying they're terrible but you know we just want to recognize part of the thing that we like in decentralization getting power hours to have actually more of this exposed to the users and the people who are creating the software and the value and then sub-optimal feature roadmaps I think the Highlight one there's a lot of them but one of the Highlight ones is you know who here is excited about trying to get software to work in Amazon and then recreating the infrastructure in azure or in Google cloud or somewhere else none of us are so decentralized protocols the ideas that you know no one owns them you you don't have a lock into one vendor where they've integrated into the stack anybody can go in take it if ever the protocol becomes sort of you know you don't like the way the Dow is running you can Fork it and restart it so the ability to kind of exit is important and you don't see features realistically you don't see features that allow for that there are vendors who try to do it but it's it's kind of like going after compatibility so billing Tech uh now that we've kind of covered all of that let's talk about the performance that's going to be needed for decentralized public cloud and I just want to say flat out you know we're not talking about hundreds thousands tens of thousands we need to be able to do billions of transactions per second there are so many API calls that are going around in the world latencies of you know hundreds of milliseconds are nice but you know the actual billing and the metering on AWS it's all handled in microseconds we need to get there too and we need the cost to be Nano nanosense and I could just say you know picosense femto sense just tiny tiny numbers it needs to be down at the cost that you know it's only the overhead of like a little bit of additional cryptography that's going along with the metering that's probably in there in your connections anyway so layer 1 blockchains we know are out they're not going to hit this they're really settlement layer there was a lot of excitement uh when Roll-Ups are discovered and they're a great way for us to scale out blockchain transactions so love them entirely we've got to use them but there's a technology and I'm going to say a little because back in 2018-19 before we had Roll-Ups there was a lot of idea about State channels being a scaling technology and then Rolex came out we didn't need State channels to scale blockchains anymore so most people in the world stopped doing State channels happily at consensus mesh we've had a r d group magmo that's kept going on with it and they've done a great job on state channels and they've improved from the state channels that you know your grandmother or father knew to a second generation that's way more significant for us so let me just reintroduce original state channels where hash time lock contracts idea very similar like Bitcoin lightning and the problem with that is that you have to exit all the time to the chain and the intermediaries that facilitate transactions are in the middle of every transaction so latency as well as scaling it's not going to get there so what actually it originated I believe for the most part in the founder of the magma group um that this idea of virtual channels and in Virtual channels you create a new layer where the base Ledger technology Ledger channel that connects to the blockchain that stays there pretty much indefinitely you have a relationship to a payment provider payment facilitator could be a decentralized party it could be a bank you could just go into this through a credit card or through crypto and when you want to do a transaction they facilitate a channel but they set up a Channel at a higher layer a virtual Channel where you're now talking directly peer-to-peer on the other end of the same TCP connection that you're doing your service with so these virtual channels get you down to microseconds and they achieve the transaction scaling because you get the thousands ten thousand scaling that you get if you're only going to your uh getting a bill once a month and settling up and the virtual channels are infinite I mean everything you do in that month is captured so the scaling is able to handle the public cloud so here I just put the names when you look at the slides you'll see sort of the API version of an incentivized API and then normally I don't try to show a complicated diagram but I want to emphasize this is a big complicated problem we're talking about scale of these payment channels being billions tens of billions of them standing I mean every server that's connected to any other microservice has one of these open across the peer relationships so I show a little example of it here but I mean it's just so much bigger than that and the virtual channels are everywhere but it's distributed through all the servers it's not going through a consensus of a blockchain so it'll scale uh it works so and then just going back from that view from the developer point of view they never want to see this when they're doing a SAS or pass project they just want to see something that works so from them what they see is an API that gets them a payment channel in addition to the one where they're doing their actual data service micro Channel and money flows One Way data service flows the other meterings included filecoin retrieval payments are an early example just want to say we have examples of that already and we should be really expanding them a lot more so I'm good I'm in good time four minutes I've got only two slides left so um where are we today if you want to check it out there's an example we did a couple years back just to sort of test out when State channels went out of Vogue with Roll-Ups could we add this to like peer-to-peer projects and do something to an existing code base so the team took the web Torrent code base and they made a service called web3torrent you can try it out it's on Gurley network if you follow the link and remember I I have all the slides are going to be on my Twitter at rjdrost so you can follow that and check it out um you know you can download a file from it I mean it's a POC thing so you just kind of try it out but if you set up a few machines download a file start serving it you'll see that there's actually incentivization and tokens that are being exchanged and that was sort of a toy example POC um luckily the people at filecoin were really intrigued with this they wanted to get to the scale on filecoin Saturn which is the very exciting web 3 CDN that they're putting in front of filecoin that gets things like ipf and some filecoin down in its initial versions down to the low 100 millisecond kind of range and in the second version we'll be able in like local GEOS to get down really to like that tens of milliseconds very high performance range obviously in a data center can be in microseconds at any point as long as you're pinned to the same data so and that's the plan this is actually coming online this year we have it working but it's going to be a lot of productization then the network has to get stood up and so forth so where do we aim to go so right now we've got the first proof points we've got the state Channel technology that can actually get to that performance it's all being done as a public good open source project the intention is that this is something that we should see all over the place as another crypto economic primitive that you can integrate into the blockchain on one side and then offer developers on the other still a lot of development to go but we're going to keep doing it this way um and uh you know we want these incentivized apis to be easy for adding to any project and um you know like really easy like the next version is an RPC Library so it abstracts away a lot of things that make it kind of complicated to deal with and then the version after that I mean there's you know things like open API API generators uh you know we're looking into can it actually be something that people who are doing you know SAS software or past software and they generate 30 different apis out of it that we can just actually have the payment abstracted away that just semantically tell open API here is the blockchain here's the payment provider Network and then here's the API technology primitive and you know like there are a lot of Open Source projects that are never going to really know about blockchain but they're going to want to run on decentralized public Cloud infrastructure we want anybody who's doing an open source project in that open core model or you know even altruistic but definitely open core models to say the decentralized public cloud is a thing that we all collectively are building the lower level iaas for and that mongodb and load balancers and orchestrators and all of the stuff above up to the software layer can run on and we can start actually you know appearing on those graphs and being a decentralized public cloud so thank you very much I'll do one shout out for the group because they're a great team I don't do any of this work I just have the pleasure of being um the head of r d so they're awesome group of guys you can see their mountain climbers from Alex and they are hiring so if you are interested in getting into this field and you like doing open source public good stuff follow or follow me right now I just need to make sure that we actually have the Job Link correct on there it may not be right after I post it there I notice it seemed to disappear but uh thank you very much all right pleasure talking to you today [Music]