Filecoin Retrieval Market Builders Demo Day, 16 Feb 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Retrieval Market Builders Demo Day - Meson Network, New Web Group & Ken Labs - February 16, 2022

Description

Every other week, the Retrieval Market Builders get together to share progress in their projects in a demo format. We want to sincerely thank all of our collaborators for demoing their developments and helping to improve the FIL Retrieval Markets one demo at a time.

February 2, 2022, Demo Day: In this video, you can find quick intros from -
• Meson Network
• New Web group

and a progress update on the Pando project from Ken Labs.
__________________________________________________________________________________________
If you would like more information about the Demo Days please follow the link below to the Demo Days planning document.

https://docs.google.com/document/d/1IrNbBm2_P79Xi2fXFj9YUfQ0PjWjreGdPYFjFntD1QM/edit?usp=sharing

A

Okay, uh welcome to this week's retrieval market demo day. Recording the date is wednesday, the 16th of february 2022.. Today we have a few new presenters presenting for us. We have sherlock from meson network, we have edmund from new web group, and then we have uh william from ken labs who we've heard from many times before, and it's always great to hear from william again um so with that said over to you, sherlock for the first presentations.

B

Sure, thank you, patrick uh thanks for having me today and nice to meet. You uh miss nice to meet you guys. So I am I'm sherlock, I'm the co-founder of the mission network and very happy to introduce the scene to our building today. Patrick share my screen.

A

Yes, of course,.

B

Okay, tell us about three yeah okay, uh so the things we're building is called mission network, and today I want to share with you the uh our explore about the building for bandwidth marketplace.

B

So, um let's start I'm a backhand programmer and I used to rent a lot of machines from like aws, almost like cloud service provider, and then I noticed okay in the end of each like most bidding cycle, I checked detail of the usage of machines. I found looks like around 30 or 40 of my resources is underutilized and some of them are bandwidth. So I try to think. um Can I find a way to monetize my underutilized bandwidth and then I do some survey. I try to find okay. Well, people need this.

B

How much will they pay for this part? And after that I found okay, like some shop video company, uh they spent around 700 million dollars per year. Only one company did spend this money for the bandwidth, so I noticed okay, so it looks like the resources is quite valuable, but then the next part I found why? Then no one just tried to reach me and integrate my resources because they are such valuable and then I found okay because lots of my resources belong to the long-term market.

B

It's quite small, so the model in the uh rhino internet world is they use. The people to like have a meeting have a call with you. You sign a contract, you make the deal to integrate resources, but if right now the resources is more distributed all around the world. So can we find a new way to aggregate these uh very long-tail but also distributed resources? And then we try to give a solution.

B

So we can regard this as a blockchain based cdn service, so we use the protocol or more specifically, we use token to incentivize our users and then let them to uh contribute their resources to our network or to our marketplace.

B

So, for example, you have some machines based in the data center idc and then you can download our. uh We call them a terminal like a software, I combine the software and you run this software, your machines and then you can serve the request from the network.

B

So this is called terminal side and because we found when you use the token model to incentivize the users, uh you don't need, like the hires lots of people to uh one by one or to integrate different resources in different market market or in countries. You can only to focus on.

B

One thing is how to build your protocol and how to make the token uh how to design a very good token model to incentivize all of the uh users or the entities in your system, and then uh you don't need when you're opening a much bigger market you don't need to hire, for example, 10 or 100 thousands of people to help you to do the video or do some marketing job so that this idea will be in things and then the next part is when you do, when you try to aggregate resources from the we call them supply side from the front like like the the nodes.

B

So one hard problem is: we need to prove that uh their resources, contributing network is working right, for example, the uh sorry because the code in the uh this network is open source and we are trying to solve a distributed system with all called open source, so people can very easily to change your code and they can easily do cheating and then, with this assumption, how can you design a system to to solve this? So we can look at the problem or solution happening in the story side like five coin. uh They use the proof.

B

They use lots of proof, steps and then to make sure that the users or the terminal nodes they it's very hard for them to do the cheating job and then, in a bandwidth part. We also try to think. Can we use the this way to solve this problem? For example, we design some proof mechanism and then use this mechanism to ensure that the resources is like we observed or working in the network. um As far as for now, we pay lots of efforts, but we still can find a very perfect solution.

B

So maybe you have some like uh other guests or other speakers. They have very good idea. uh We have lovely to hear from them, but fun for us. For now we can't find a very good solution to do the proof, jobs and then, uh when we're talking with my uh team member, we think okay can we think different. uh It's like if you're doing the proof it's hard. Can we can we do another side like we. We do. We don't care uh the how much data you have transferred right.

B

We just let the network to measure your uh resources, so this is things having our system right now and we're all still in development. But we can share some idea with you, so we call those retesters and then on the left side, there are just a normal minor nodes and then in the right part you will have some speed testers.

B

We don't know where are they and the job they do? Is they will randomly select and test some nodes to get their parameters and if the node doesn't respond very uh not accordingly, for example, we have two or three testers, they uh test the same nodes, and then they have a different result.

B

So during that time we can have assumptions that maybe some nodes is doing cheating job and then, for example, in the around a one hour later, when we test again the same node and the error capacity is, for example, only half of the of the a number before so we can assume that they made doing some cheating job. So this is the first exam called speed.

B

Testers like here, you can see if the speed drops a lot when two nodes tested at the same time like they have the correlations, so some notes gonna be penalty, and then the next part is how to design our algorithm. So at least we have two models to solve this problem. First, it's like usage model just this strategy, so reward is based on usage like filecoin.

B

Unlike many, for example, design is in the story part when you store the files, you're gonna, the more you save the more you're gonna be rewarded right so like in this system, the more data transferred, you're gonna be rewarded more.

B

This is first strategy and then the second strategy is, we only measure your capacity right, so we don't care um how much data, how much real data you transferred, because if you transfer much more, you will still get the same reward, and this idea is just like we like uh to think about things having a serious life. They have the base reward and then they have the transaction fee. So the ranking renting model is much like you will have the base reward and then the usage model is like.

B

You can add something for the transaction fee right now. We just try to solve the problem of the rating part, the the basic capacity- and this is mission used and for the real usage we're going to have two scores. The first is called bandwidth score and the second is called storage score because we know for the bandwidth system or we can call an obsidian.

B

uh You also need the ban, the storage right, but the difference is, you can have maybe uh several gigabytes, for example, one unit of the storage, and then you can compare it with 100 units of the bandwidth part, so just like stored once and be retrievable for a million times, and then the multiplier of these two scores will be our final score for the manners.

B

Then the next super hard problem is still in challenge. The whole system is, you need to solve at least like traffic router right, so the users based in different locations, your geography locations and the nodes also will be in different countries in different locations.

B

So how can you use a system or a strategy to find a very good router for all of the requests right right now, we'll just make it much more simple right. You have the iv geography location and then we can have http 02 to redirect to some nodes and then before then we can build something called load balancer to cover the quest, but I think we can maybe we can with the things uh it rates. We can find a better solution for this part and then the next part is called schedule system. So it's yeah.

B

You know about the pictures, uh so this is super hard problem. I think in the uh way to match the demand side and the supply side, so we can think this model as the same thing like in the in the uber right, the uber will have the their uh drivers. They have a lot of drivers, the the taxi anything else, and then they will also have some clients.

B

So, for example, if you have two clients in this place and then you have two cars near these two guests, so the best schedule system or the algorithm is not let the the most near car to deliver you right so because it will not get the whole best result. So this is called schedule system and it's qualified things happening in the uber side, and this is also a things we try to use.

B

We use a very simple model to solve this problem, but then I think, maybe in the future, some people or some scientists or some very good team- they can propose a much better model and then we can use them to integrate into our system um and then the next superpower challenges is the token model. So when we design the like cdn service or the bandwidth system or vendors marketplace, we call them uh you're going to have supply and demand side. So for supplies, that's quite easy.

B

You can use your native token to incentivize for your miners, but then for the month side it will be a very hard problem because if you use like it chili's token something like ethereum and then the difference is when you're using the ethereum people, don't know how the infrastructure costs like when you use one percent uh resources from ethereum in in the real world. We don't know how these resources cost, so people can give a very crazy number. They say wow.

B

This is what computer this is the world virtual machine, so I can think the the resources can cause 100 times than the real resources, but in the cdn area it will be quite different. So people know how the like story cost how the uh like the the specific traffic cost. So when you try to prove them a new system, but with using the crypto native utility token model, it will make them confused. Okay.

B

Today I pay for traffic with like 10 units of money, but why, tomorrow I need to pay for your 50 units number right. So this is a problem existed to solving the the the system which has the target in the real world, and then the next is.

B

If you want to satisfy your users, your demand side, much better, you should keep the price stable or you should keep the payment. Token be very, for example, keep the same as like the stable coin. They can pay in the usd usd and then above them.

B

When you want to reach the very open uterus, you should say: okay, you can paint different like us dollar, like uh gdp or like rmb or different currencies, you can just support them and then, in our design uh we found. We don't want to sacrifice these two things. So for your own system, you should you should still keep the utility token right. You should only win token and then they're going to be incentivized for two sides and then for the demand side.

B

We should add some like plug-in to help them like they can pay in the their the token they own or anything that they want, and then we're gonna have some like some third-party or middleware to convert it to the our native token, something like bundler for our wii yeah, and I also want to show you some demand side for part. So we which we have served like a mosque, they can. They have a checkbox here.

B

They can use our service and also we help like weave to help them to solve problems. The the file stored in the salon, art the the project called dj and they use rwev as a storage system, and we help them and third is: this- is a benchmark page about the missing and ipad fast, so the smaller is better. So we have we do some benchmark and we find that we have a looks like a much better benchmark, result yeah and welcome people to test for our service and then the next part is.

B

We also try to compare with the right now. Storage system, like our wave so uh also the same thing, gonna happen in the five coin. So, like storage system, they have incentivized for the storage and then maybe we can use different systems or different networks. You incentivize for retrieval, and then we can combine these two uh system together to serve you. The users, for example users just want to think okay, I just want to store the file and then I want I just want to consume the files very quickly.

B

So, like things happening, um sam started send they said: okay, jc's new 500 megabytes empty stored in our wave and their level by this network uh yeah. This is the last page. Okay, uh you can try mission so right now we are still in testnet. uh We welcome people to test our service and to give us more advice and also we we still uh try to solve those the the challenges I I mentioned before so like when you have the url.

B

uh You add these, you check your audio, you add you're gonna have the original I o and then you're gonna have the city and io, and then you can use this series. uh Okay, guys thanks. That's my presentation.

A

Nice thanks so much sherlock um that was great um and yeah. Please reach out to sherlock anyone's got any more questions about about meson. um The next speaker, we've got edmund from new web group. New web group have just started a grant um as part of the retrieval markets, working group and so yeah. Looking forward to this introduction presentation over to you edmond.

C

Hi hi everyone, uh my name is edmund, I'm on behalf of new web group, I'm responsible for overseas communication and translation with like uh to communicate with global communities, including portugal, labs et cetera and as our international office just established in august 2021 now with like over 50 people's overall and secondly, the new web group is a starter service provider that focus on distributed storage services and hardware, researches and development, cluster storage and idc computer rooms, hosting etc. In cloud clustering operations currently has operations in hong kong, singapore and aiming for global coverage.

C

Our vision is to build the future work3 to provide with the customers, visitor safety, technology, morality and life happiness. uh I'll. Just give a brief rundown from our company's history.

C

The funding team participated in blockchain and crypto in 2016 in may, and in 2017 september they started to participate in ipfs and rebuild the technology team and uh in tucson in 2019 october, the cooperations, with local adc operators and in 2010 august to participate in fuel and space raising in also the midnight online and with a speed, update research and development and in 2020 december they established the new web group technology company uh in 2021 in march because of android underground investment from qf-capital, with a valuation of 200 200 rmbs.

C

And in 2021 august we have the new uh international offices to start doing operations overseas.

C

uh The executive team, including the troy knee the ceo with background in real estate, and consulting and he's an early investor in crypto, and he had degrees in mechanical engineering and the co-founder of a tfs, a fintech organization in grc region, also, the diamond juan the vp for incorrect qf capital specializing in ipos, mma and restructuring, and also the axis o young.

C

The cmos the x ipfs mean director and the go crypto exchange cmos and the cto is gallon tang and he was a phd in stanford and worked for leading companies in townsend and michael young, the president of international division, the earliest advocates for btc from bc, canada and leon due the vp, the director in uh in cloud storage, cluster and unco-founder of int supply chain, and also we have the the key tech team, uh the dean, the chief programmer and mr yuan, the tech vp, the both have many years experience in software development and for game for blockchain from software languages, including rust, gold, land and club class etc.

C

uh So uh I just because I'm not a thai guy, uh so I just gave a just rundown of the uh uh of their like uh translated version of uh what's your development. If I see anything wrong, maybe you can refer to the tech team and they can give you a more specific uh explanation.

C

So, basically, uh first of all, like let's take example of recent ifts there's. No, this is not a nft project, say uh in the falcon ecosystem. There's actually can uh actual stall act doing actual storage, because there are a lot of obstacles to to get on with. So in order to solve other problems, maybe we have four objectives to achieve. The first one is the need to solve the problems of real-time access and that that is all location of network resources.

C

This number two, the latency minus uh the the the the the original magnus. The revenues cannot be affected for them because they are doing what they do before as before. So is. The incentives is better to function in the new systems number three, the compared to centralized storage and otherwise providers.

C

The firepoint should be able to provide competitive service with low costs, better experience and high quality services.

C

The number four we need to do solve is to use ipfs protocols as a call to support and solve the problems to be more efficient, and this is to be more native, so they come together. So should we introducing the new incentive programs, we suggest that the future miners can have divide into three categories: three. They have different three different roles as a b and c.

C

For example, like a the existing traditional falcon storage service providers, their incentive mechanism should be to obtain falcoid rewards by providing efficient storage space resources, their responsibilities, including maintain mainly to to save multiple copies of data from miner b and to customers to encapsulate encapsulation and doing offsite permanent backups.

C

So the plan is to gradually optimize and improve the utilization rate of hardware storage, such as replacing the ccs sectors of of 14 eb's bandwidth, ep storage with real and valid datas. First, the miner b, so the second manual. So so the ipfs nodes builders can get few rewards.

C

We can we call it miner b, so it's deployed as ipfs nodes, so that removes the fail coin encapsulation process by storing data under the enterprise chain. For example, nfts customers only need to upload to get the ipf as harsh or harsh of the of the received content. So you can achieve real-time access to data's.

C

The data is then stored on top of five to eight a minus in the next three to say five days. The reward mechanism should be like referring to the falcon white paper.

C

The traditional storage rewards model design according to the storage resources provided by miner b situation and the bandwidth, qualities and bandwidth size and networking traffic actor usage, et cetera, et cetera, comprehensive rewards falcon to minor b's and his responsibilities should including ipf's, ipfs storage, hardware and network resources are provided and after the dht perform, data addressing the data will be read and relayed through the nearest neighboring nodes with a logical distance.

C

According to real-time data, is guaranteed so as to ensure a quality experience from customers to access files.

C

The the plan for minor b is they can seamlessly access, verb, 3, storage and ft storage in the applications developing by protocol labs through the ports of all storage miners on the falcon equals ecosystem.

C

So basically, the miner a's doing the copies, making the different copies and the manner b is, is trying to speed and to make the bandwidth small to get to speed up the upload and download speed.

C

So the miner c will deploy an ipfs lightweight nodes through an intelligent scheduling systems and personal dependence and storage resources are composed into light nodes to to release social idle bandwidth resources through greatly reducing cdn costs and providing low cost high quality super cd services for the enterprise.

C

The the incentive mechanism should be like aggregating the bandwidth resources of services providers and intelligently scheduling the demand outside of the services, accelerating the completion of the original algorithms of of settlement of traffic costs, plus release the demand side of failed tokens incentive as a shared cd and user fees. The responsibilities should be including to provide high quality and low cost accelerations for minor b and minor a to retry to retrieve the data.

C

So we we make a b c as a congo to consum, so they can come. They can support a very complete. The super shared cdn concept, which can provide high quality and low costs. Acceleration services for data science and short videos live stream games and downloads et cetera.

C

So we want to do this to implement the a b and c combination so that the fail coin can have a the function of ipfs and combine the the fail coin with ipfs together.

C

So what are the benefits of doing doing this with fail coin? So the proposal combined the concept of shared with idle home bandwidth resources through technology implementations and proposed the concepts of fail coin shared super cdns to the high performance cdns, so for the individual user, so for users with either bandwidth resources to make idle bandwidth resources more valuable. So you can use all those free resources available for corporate users for the enterprises.

C

They can suffer from expensive bandwidth costs and can help enterprise to reduce their bandwidth costs and free up more more funds for product developments, operations, promotions and other kind of stuff for falcon ecosystem. It can provide high quality and low-cost shared edge computing services like iots, and propose the new concept of miners in improving the the roles of eco and the miners and the reward ipf system providers to make use of the user experience much better.

C

The the project code is completely open source and all the projects are developed based on fail coin. uh It will have a high quality user experience after the completion of the whole projects and help the failcoin to uh to to to make the consensus to make value more the market more valuable.

C

So the shared cdn concept for failcoin to further achieve the carbon neutral neutrality. It can improve the inherent image of energy consumption formed by the blockchain project like btc and ethereum, and truly runs through the concepts of responding to fuel grain environmental protection. So the fel coin can achieve a carbon neutral neutrality and realize the carbon negative initiatives in the time.

C

So that's basically our idea. So I finished.

A

Thank you very.

C

A

um Yeah, it's it's.

C

A

Maybe uh maybe.

C

We will provide uh pppt after that. Maybe you can have a better idea of what's happening and maybe give give me a new link and to download the ppe so so, unfortunately, we don't have it ready for here at the moment. So maybe later we can have a more detailed explanation on that. Yeah.

A

Yeah, that would be great um yes,.

D

Please share whenever you're.

A

Yeah, it was very interesting to hear just you speaking through it and the different uh different layers of providers uh a b and c and how they how they work together. So, thank you very much, I'm looking forward to seeing how this.

B

A

Sure, great um and now to the final presentation, um william from ken labs.

D

Thanks, patrick the casino screen: yes, okay, uh so we just had a break in china. One week break it's a public holiday the chinese new year, so we don't have too much update, but we have almost wrap up the pando service, so I'm going to give a bi weekly update, so it's the 16th of february 2022.

D

uh So what's new, so I want to announce that we have updated our provider metadata specification.

D

We have some discussions with the product labs contact massey, so he gave us a really valuable review, so the specification has been updated. I'm going to go through shortly, so, what's required. Also, we updated handle topic considering the the future feature, development or the different plan on the roadmap. So we consider to versioning the topic.

D

The last thing is the updated provider integration process.

D

So let me go to the documentation so so this is uh the repo for the handle service.

D

Okay, you can see here. So what I want to emphasize is the the change of the integration, so as a provider, what we need to do- and uh so this is prerequisite. This is part of the method specification.

D

Any provider need to publish their metadata with the required iprd nodes provider and the signature, so we need to know the provider and the the whole package. The data need to be signed with the private key, so we need to have the signature, so this is. These are two required nodes in the ipod data other than that we accept also a previous id which are linked to the previous metadata.

D

So if you want to link all the data in the chain, so you can use this node, so the panel service will also recursively sync or fetch the data, the metadata. So once you once handle, handle service and there's a couple of uh metadata update in between, so it can still catch up in the future.

D

So this is a metadata specification. We also updated the topic to, as I just mentioned, pando v0.0.1.

D

So this is a topic for now we are using, and the last thing is the integration process.

D

We have tricked a bit of the flow, so you may notice the change if you visit before so I move the step, uh building, bundle, service and client to the first one, as we need the pando client to do the registration job for the provider in the future. We may have a easier way for the provider, but for now we have to do it in this way. So we build the client, go to the brain directory, use the handle client to register itself with the peer id and the private key.

D

So the priority key is guaranteed. It's guaranteed that private key won't be used other than just the signature, so it won't be shared or uploaded to any network. So also the code is uh open source. So we can, we can review so there should be no risk.

D

The other thing is that we we create a handle sdk. I have a example here, so we can go and see it shortly. So we use sdk to connect to handle initialize or build your ipld metadata. We call the metadata instance, so you can push the metadata.

D

The last step you need to do is execute update to root. So it's not explicitly mentioned here. So this is part of the codex component, so this is how we uh publish the iprd data to the topic: bundle 0.0.1, okay for the rest of the document, therefore, the consumers, so how we can access the metadata or snapshot status, how we can access over a graphql api.

D

The most important thing is how we can access ipod data. So it's following the golex component specifications, so we many consumers can also use the same golex to fetch the data.

D

That's the integration process in terms of the example.

D

So we can see here so this is an example of the publish, so we have a panda sdk for the providers, so any provider can use it to easily connect to handle then build the metadata, and that here with the publish.

D

And the push so we got the metadata cid.

C

D

So this is the example how we use the panel sdk.

D

So if anybody interested you can go and check out the document here and if you want to start a panda server by yourself, you can follow the steps here. We are using. S3 gateway for the uh archiving, so metadata will be archived regularly and uploaded to varco network.

D

So for now we are using this gateway so to use it and run the server on your own, so you need to uh register and get your extra api key and set it in the configuration file. It's all mentioned here.

D

Okay, I think that's it for today. Lastly, uh the this is the contact information of the can labs so feel free to reach out. If you have any question or any interest thanks.

A

Awesome thanks very much um cool. Well I'll. Stop the recording there thanks. Everyone.