DePIN: Pushing Decentralization Beyond Blockchain by Akshit Gupta | Devcon SEA

[Music] everyone am I audible at the back yeah okay awesome so hi everybody I'm akit I lead the engineering team at huddle 01 and we are a deepin dedicated to making realtime communication affordable cheaper and performant so uh we are a deepin project so if you're familiar with deepin deepin stands for decentralized physical infrastructure and uh we saw a gap in the market that uh real-time communication is very hard and it's uh it's very pricey as well so people like Zoom like companies like Zoom Google meet all these rely on heavy infrastructure and uh have to host it on cloud providers like AWS so this is what my talk is about how deepin can make it cheaper so as I said deepin stands for for this decentralized physical infrastructure and in blockchain we've seen a couple of Trends uh from the start things like nfts have shaped like reshaped ownership digital ownership in games you have these skins and everything and these can be soed via nfts uh to make sure that uh only one person owns it and stuff and smart contracts have automated uh money transfer as well so you can program how smart contracts transfer for money on a based on based on a particular condition so but we're still deploying these things on centralized infrastructure like AWS so if you were to build a complex app which requires heavy infra you still host it on AWS things like storage and bandwidth still require centralized infrastructure so that's where deepin comes in uh we saw this uh with filecoin so filecoin is a really good example of a deepin project which is decentralizing storage so similar to filecoin we have uh it for realtime communication so uh charging for stocks and paying for flows so this is a very interesting Trend we saw and AWS was charging crazy amounts of money for egress costs so as you can see in the slides uh AWS charges like 700x amount of pricing for egress costs so say like you fill a bucket full of water water being the amount of storage then the the amount of water that's being transferred out of AWS which is data being transferred out they charge like really costly amounts for that and not for the stocks that they save so that's how uh AWS is charging and it's like 700 markup so I am bullish on bandwidth projects like dpnr so internally at hdle 01 we did an interesting study where we calculated the bandwidth costs for hosting on AWS like hosting a real-time communication app on AWS and as you can see for a single participant the bit rate consumption was around 700 uh kbps and for eight speakers in a room and 20 listeners the data consumed was around 112 MVPs and if we were to take like 100 such rooms uh hosted on a particular uh app you'll see that it's it comes around 11,200 MVPs so data consumed in an hour would be around 40,000 gbits okay so that's a huge amount and if you were to calculate the data transfer cost on AWS that comes around to $400 okay so if we were to use a deepen project and these Services were hosted on a physical infra where people are incentivized to run these notes then you the price goes much lower okay so these are some popular deepin projects you can see huddle 01 which is uh the project we we are working on and some popular projects like grass are also working in this bandwidth department where they're saving cost all you need to do is just install an extension and you'll be good to go so if you are looking to build a deepin the best way would be to look for a market which is super overpriced and you can easily save costs by um allocating that job to a particular person who can run it on their own machines or uh host it somewhere maybe with less P so as you can see the less spe the better this is one thing to follow if you are looking to launch a deepin project very easily and uh it's very easy to uh deploy a project with less p as there's no physical part involved there are really good projects with more p as well as you can see on the left uh but with less P you'll be able to run it and uh make it work so thank you that was a small lightning talk uh by me and I am happy to entertain some questions thank you okay okay do we have any questions from the audience anyone let's wait for another 5 Seconds okay oh do you want to throw [Laughter] it yes perfect so you talked about uh essentially that the less the Fe the better you'll need to put it it's a mic hello hello oh wow yeah I know so you talked about a point where essentially uh the less be the better I just wanted to understand if you have less spe there's also a chance of having uh more chance of gaming the system that means verification can become uh tricky if your system is more digital rather than having a physical Network for example helium it's very difficult to game a helium system because you have a bobcat minor uh but in case of something like huddle it's more of a digital Network that means you require verification systems more like qos systems so have you thought about that at hard one where you can have a qos system implemented which can act as a similar to a physical infrastructure Network as well so that nobody games it yes so uh the best way to go about is to First solve distribution as we did because the main concern is to save costs because uh people will only come if you save costs but if you are to move ahead and look it in the long term then you'll have to come up with a ZK system which is provable in nature so we have to look for a way to make the compute uh verifiable provable using things like ZK as uh they've come and right now the way to prove like make it provable is very difficult as things like circom have their limitations as well but the last step would be to make it provable in some sense yeah awesome I think if you're able to do that then communication becoming provable in a digital environment will make it a highly scalable environment and make communication much more performant and scalable thank you so much yeah thank you thank you anyone else okay one over here sorry we have only time for one more yeah H look said amazing talk by the way so I got curious to know let's suppose if a company like Google or Microsoft want to adopt a deepend do you think it's that scalable right now or maybe in future so uh if Google were to do it then first of all they are in direct competition with the deepend project as they have gcp and uh they're charging like egregious costs for those and so it would be difficult for them to adopt but eventually once like uh the deep in Project scales and people start adopting it then they'll have no choice to you know move on to these cheaper options because it'll be much cheaper for them as well okay thank you