Build and Operate Internet-Scale dApps on Celer Network

Transcript

all right as everyone starts making their way in I'm about to introduce the stage mo dong from cellar Network actually privilege of winning a bounty at East Denver with Sela and ma dong is actually indirectly paid for my rent last month so I'd like to welcome on stage mo dong a round of applause please he's gonna talk about building and operating internet scaled apps on the cellar Network alright hello everyone I'm mo from cellar and today my I'm gonna talk about towards building and

operating internet scale decentralized applications on top of Saturn Network so this is gonna be a short talk so what I'm gonna touch base on is first of all a very quick introduction about Saturday the technology and also a user product that it will release the call center acts and then I'm gonna talk about you know the testosterones that we did during the last six months and based on the data as we gathered and the user behaviors we observe that during this entire task that process we will talk about why developers should actually care about building on top of the seller and also some very valuable lesson we think we learned from running this world's largest scale generalized a

state channel and a layer too often scaling Network and finally I'm gonna briefly touch base on you know what snacks for setter so what is center network center network is this layer to scaling architecture that is like a you know I bet everyone knows sore for like what layer two is but certain network is proposing a coherent object scaling architecture here with both a technology stack and also a crypto economic stack on layer 2 so on the technology front we proposed actually a layered structure here with the lowest layer called C channel which is solving the problem to support generalize the option scaling for not only just a payment but also arbitraries smart contracts and in the

middle we have 0 because the seller network underlaying is always running and operating a generalized offering payment network and the seer arts job is to make the you know value transfer extremely efficient and in fact prove what a optimal and then on very top is basically our SDK and the user facing layer called cos and you know and this this layer manifests itself in terms of seller acts the application that we build and besides the technology architecture we also propose that the first a layer to crypto economic constructor called a signor me with the three part why's that called proof of liquidity committing the mining and the liquidity backing auction these two part combined actually solved the

liquidity shortage problem the faced by a lot of state channel networks and layer to scaling structures and the state Guardian network is proposed to solve the so called watch tower problem in efficient and you know a reliable way using a size gene constructor with the same spirit of plasma so I wouldn't have time to dig into details about each of this aspect but many of this stuff is are already available online especially on the crypto economic side if you check on the center networks YouTube channel there's a narrated talk about this part in pretty detailed so we actually released some developer tools already we

have iOS Android and also web SDK during the past few months so we bring in a different hack zone and also collaborating with developers to build games micro payment services layer 2 device like layer 2 prediction market and also messengers are integrating with like payment master micro payment earnings in the payment like 1t means East Amer actually build a micro payment solution in by integrating center with telegram so based on the SDK based on the Mobile SDK we actually released something called seller acts or a test not the seller acts is really a user facing access point to Center Network for the user to use Center and use the

application build on top of seller so it allows user to interact instantly with each other on blockchain it it will probably it will bring a lot of things for you to play around and finally the core you know value proposal for the user is always you can spend less and way more and are more in up on the blockchain so this is just kind of like a quick demo of seller acts that we I just recorded last night so it looks like wallet but the difference here is that you have the block construction then you have the seller pays action what the set up a section is really or off balance or layer to balance here now what I'm doing now is like just sending an option transaction now just like that the option transaction is done it's instant final you know and the token we

use is actually er say 20 token generates down robson and you can replace that with any kind of a stable token of your choice when we launch I mean that now what's happening here is that actually like we're trying to play a state channel game with each other where one guys take let's say $1 and the other guy also like choose to take one daughter and they created a basically a game and when they join a game was what actually happened on the lane is that both of them send each other offering a conditional payment like think about lightning is basically sending a unconditional payment of change but what sir can do is that their can send you conditional payment based on condition unlike this games result

now what is this game this game is also fully running in a of king stage channel like this is a generalized say channel in action right so you can see you can already kind of a see the certain networks a interactiveness is already proven by this kind of a very short demo that you can actually play that game extremely efficiently and interactively on blockchain so doing imparts the six months we release the two tests not one called Centauri we released around October it was on 18 and the other is series so we have a huge amount of you know teachers already supported including multi know the non coaster image analyzed payment networks whisper both the agent years

twenty whisper generalized a channel with support function or an imperative stage hog contract interfaces option on chin message states protocols we actually wrote a stew protobuf parser to realize those kind of things and the list goes on but the important question to ask is why should any of you care right so like sarah has all these features why should anybody care well this is why you should care from day one to day 172 we grow the user from the first among sis monthly active users on the test not from 310 to 18,000 you know just easy in the beginning of April so this is the 56 growth in five

months and we believe that we are operating the largest generalize the state channel network in the world right now I don't think there is any debating that so the users are from 67 different countries as we can see from our back-end databases and interestingly you know if we look at Madagascar the Madagascar actually have a very high active user to population ratio so maybe we're like host and meet up there someday so it's not only about the user and geo locations of this application of the rich but you know we are we also are seeing extremely encouraging you know product market fit or you know growth matrix that makes we think that certain

actual technology actually can fix many critical pin point of today's decentralized application like low retention you know hard to get on board and all that stuff so for example it's not like you know this eighteen thousand eight users just come to a certain network on board and then never comes back they actually on average plays ten games monthly and we have about four million transactions processed on the test network mostly and for every active user per day they spend about 12 minutes you know on their daily user engagement time even for traditional application traditional mobile application this is already a pretty pretty high number and we already achieved that 20 percent they seven retention ratio so if you look at

the conversion look at the retention look at the stickiness it resembles not a decentralized application it resembles a centralized application and I think that's where we are trying to go as a community so it's also great to note that this is just one application on a paper money that has not and we believe that with this kind of a smooth interaction enabled by seller and this extremely high transaction volume involved in certain network is extremely easy to create a very sustainable monetization path for developers uncertain Network and that is I think why you know I was suggest you to keep an eye on us as we grow mature more and more mature in

terms of the SDK and developer documentation and everything of course not everything is like awesome in the flowery there are also interesting lessons learned during the process I don't have a lot of time so I'm just going to touch on some important points or interesting point rather so this is the same graph that plotted the Engel Law Group is maker scale there are some key events that drives the growth so the first one is like in San Francisco we launched the stuff and then we launched around like a Chinese community so that you can see there is another pump and then we did like a Christmas tournament and you can see another small bomb here and then we release the stellar ax which is like the application can download it today and then we have e stammer there's

The Grocer kind of a sustained and then you know the most recent jump is because we launched on bi'ness launchpad so the message goes out and then you know with the user just jumped so at each stage there are all some very interesting stuff but I'm gonna like to focus on several of them so the first part notice this drop here like you know kind of a we released the which had you know message or like kind of a reach has a small promotion to tell user about zero acts and then suddenly it's start to drop well so the lesson we learned there is that you know I think it generally applies to every DF that is a you know it's kind of a hard to do global reliable event monitoring in the

beginning we rely heavily on in Fuhrer so you know we thought okay this is probably fine we deployed the application also on in China and then like you know because we open the channel and there's event imitating image channel opening car and the mobile application would catch that channel winning car in the beginning it works okay but because Akira was hosted in you know a table s and also use this kind of TCP connection and WebSocket stuff it looks very much like a VPN basically so you know we're starting to see very very interesting connectivity issues that coming from China and at one point in Tarzana went black so basically we grow a user we return it back so that's just a kind of

a hard lesson we learn and we did something to leave it in the matter you know contract that another thing that is kind of interesting is that this just kind of a weird shaped the same so what happened there is that we all like talk about theoretical a you could have this scenario where like kind of a you have a lot of state channel and all of the state channel won't wants to start all at once and then there's some like people just cannot like a set of the channel properly and all that stuff this is a water people call sediment storm this thing actually happened to us not because of a malicious user but because of our thought in our software so during the upgrade process from the previous

version of test night to the new version of test net there was a backward compatibility show we didn't catch so every you know old version of the app is actually trying to settle the channel with a you know order of malicious you know state proof so we kind of a self-created attack to the network ourselves but the problem about this is you know at the time we even though we did some even though it is something like a mitigation effort you're kind of a prevent this our offering service provider or like the hub's actually we acted the correctly and trying to reset all the channel in a cracked assess but because like the blog was totally congested at during its the

mall period of time and because it's attached nervous at a very small amount of time out for the settlement it'll become actually kind of a trigger this kind of a settlement a storm so the last thing we learned there is that we really need a origin Oracle to tell us what several blocks like before this is actually congested in or not so you know in the production lounge will incorporate the such article but we would hope to have layer once the porous some a primitive to tell us this stuff is happening that blocks the congestion so you know be careful about this stuff and you know we are of course like after the bynars launch pad you know announcement we see a channel creation storm literally like thousands of the

people trying to open state channel at the same time without offering service providers it created like a big issue in terms of event monitoring because the event more into me and the TCP connection does have timeout and your phone does is not like actually very reliable its convert become like another event monitoring you know issue and that also caused us to turn away a lot of users because users are actually confused about what was going on why Cannot you know get onto a setter so we didn't prevent user or tell user what's happening up front and that kind of a fight us we actually like the highest rate we turn a turn turn away the user was like more than fifty percent usually turn away actually

so now let's finish thing talking about some technical stuff let's talk about the u.s. so these are the UX comments we return we gather from East Amer you know really nice to use a great layer to experience play it game on that and all that stuff serve us is awesome well not until I start to push the stuff to the more general public and then I realize this is credible we're leaving a bubble situation so so because II stammers the crowd is actually very interesting crowd in the sense that I you know everyone knows what layer tweets and never one kind of expected what you have using layer two but when we launched this

stuff on the Bynum's launch pad and let everyone know about this stuff people just generally curious about the blocking start to download server and start to try to use that stuff and then we realize layer to you axis actually a trickier than layer what UX some tough questions we received from the users the the biggest question like the most frequent last is probably the first one you said are just a centralized server holding my money like his exchange the reason that they ask this question is extremely legitimate because there is always this process to separate the ancient balance and offering balance right so you always have like your own team balance that Euler sort of already understand that is a decentralized that is safe that is in

his control but once somehow you if you want to accelerate and use the application build on top of setter you need to move your own team balances from your own gene wallet it uses optional word this moving process feels like deposit because you are also sending this stuff you know why our transaction so it takes a lot of explanation to tell you that this is actually knows what's happening but still like we're kind of also still struggling with that and trying to come up with good analogies or good ways to explain that to the user and there are things like okay I joined the center but I cannot send money to my friend who didn't draw in Sarah what's going on and you know there are constraint like where did money why didn't my money call stuck when the

player I'm playing with just laughter the game well you know because of the anjing settlement and the layer - you know settlement process even if no one is kind of a you know dis building in this process it still takes a timeout to let the of ching pending conditional payment resolve back to a unconditional non-payment and these kind of stuff we need to explain to choose a user who does not even fully understand layer to layer what not even to say about layer two so you know that I probably don't have time so I'm just one briefly touch on what's next for seller so for seller we're thinking about moving forward towards a coherent layer to platform that actually utilized different the

technology pieces not only just the state channel but we actually in this state Guardian s we're already using you know plasma like site kings and also we think that you know the boundary between siding and state channel may not be the strict but that's what we're going like we were using utilizing different technology pieces with and mass adoption rather you acts and when may not mean that is coming in q2 2019 and with all the expiry we gathered from the test neurons mm I runs we are we just launched a new private alpha for you know to prepare for our meeting at lunch today so if you're interested in like scan this QR code or go to bitly /center XRF our

registration and you know we'll invite you guys to the private alpha so is that conclude my talk thank you [Applause]