Filecoin Retrieval Market Builders Demo Day, 16 Mar 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Retrieval Market Builders Demo Day - KEN Labs & CryptoNetLab - March 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.

March 16, 2022, Demo Day: In this video, you can find quick demos from -
• KEN Labs
• CryptoNetLab

A

Okay, welcome to today's retrieval market demo day. The date is wednesday. The 16th of march 2022 and today we'll be hearing from william from ken labs on joust db and then hopefully nicola from cryptonet lab about some of the latest stuff they've been working on. uh Please put any comments or questions in the file coin.

A

Retrieval market slack channel and over to you, william.

B

Thanks so let me share my screen.

B

Okay, can I see my screen? Yes, okay, uh hello, everyone. uh This is william from ken laps, so I'm going to introduce you the child's db, the rpod ledger database, which will be scaling. The blockchains with the option trusted data storage from kendall, so I'll, take about 10 to 15 minutes to introduce you what the glc db is and how you can benefit from this solution.

B

Firstly, let's take a look at so let me remove this tiny window. So let's take a look at the the architecture, high level architecture of the gel's db and how we envision the the future and solve the problems in the usual blockchain scenarios.

B

So uh can labs division the jobs family to explore and explore how components such as record keeping this is about storage and smart contracts. This is about computation can be moved off chain to enable more robust, computation and storage requirement, without sacrificing security and scalability requirements, for example with option storage.

B

The data known no longer need to be hosted by all the nodes in the network, but just the nodes that are performing the computation, so another example is so computation engine so doesn't need to be packed to the layer one chain instead, so the computation can be performed as an option node so, which you already see here. The option node will do the computation job.

B

So in this way we abstract away the storage to the option provider, which means we can immediately reduce the redundancy storage requirement.

B

So in this scenario, when the transaction is received by the channel, the node must fetch the initial state. So there is an initialization phase. Initial state from the storage node after the computation, the chain node will send the result or the resulting state to storage to persist.

B

So in this way this is how the computation can be performed by the object. Node.

B

In similar, so how the storage can be so this is the scenario we how we can abstract the storage to the option provider or the node.

B

The computation can also be offloaded to a similar node, so we can have an option computation node. So in this way the state transition function can be computed by a separate computation node, which is the option. Node again, we didn't show it here, but it's a pretty similar concept so again. In this way, we don't need to store the computation input and the output on all the nodes. It will only be needed by the option node, so the redundancy, the storage redundancy, is not needed anymore on the whole network.

B

So this is how we design the the whole option node to the uh on-chain network.

B

So, as you can see here, we in the king labs uh java family, we will have the judge db, which has a spark that would be then from the name. It's pretty straightforward. It's a high performance storage solution and in this case we we can say it gives you the key value storage to to the state storage and the the jobs spark is pretty similar from the name. You if you are familiar with, like apache, sparks, give you the competition.

B

So in this way we can power the off-chain computation node. So this is how we uh design the computation and the storage with our program. Family.

B

Okay, so this is a revolution to your option model and, uh and the gel db product is to scale the blockchain. With the options trusted data storage.

B

It can give the web3 ability to to hold the the information or to manage the information explosion so that we withd no redundancy data can be better managed and processed by the option. Nodes, including storage and the computation.

B

And this is how we, how the gel db can solve the data integrity problem.

B

The other thing I want to show you is the the brief or the high-level two-year roadmap.

B

So in 2022 we will start our architecture design from in april and by end of october. We hope to see the storage layer and storage and the application layer ready for uh usage.

B

uh There's, there's probably no child db, the first version of job db, yet the commode components, the key components should be ready by october, so we can probably we can start using them in some scenarios as separate modules.

B

But anyway, by end of december, we we will release the first version with the key features and the interface to the users.

B

By december. Probably we don't have the sql support, which will be essential to web 2 businesses, but in web 3. We hope gel to db version. 0.1 can start be start to be used by the web 3 scenarios in some projects and in 2023.

B

By end of february, we will give the sql support under the javaspark for computation capability.

B

This is uh the time when we can start serving the with two businesses, so they can use the sql interface to read and write data from judge db and manage the data or even do the computation.

B

uh In july next year, we plan to rule out the the cloud service, so this is a a place where you can subscribe the service without having to manage the dell db network by yourself, okay. So this is a the roadmap.

B

This is the javadb family, the ledger database kv database and the spark the computation platform.

B

The last thing I'd like to to share is this: one.

B

Okay, so this is computing over ipld, so you can see here. We have multiple transactions going to the chain node and the computation will be done as tasks by workers so that how the workers are scheduled. It's by the the jobs orchestrator, it's a scheduler. It will schedule the tasks to different workers based on specific on certain policies, so talks are done by workers.

B

The necessary data storage is down to this part. The storage node and the result will be synchronized back to to the chain node by the uh pops up mechanism so that the node will know when the computation is done and the where it can to it can go and fetch a result. So this is the the.

B

High-Level architecture of the computation, so because the computation is so in this architecture you can see the computation is decoupled from the blockchain node. That means the computation can be wrong like a serverless function. So like the typical.

B

Fas fas so like the amazon lambda service, so the option competition allows a combination of the the blockchain ledger maintained status of the state with a serverless function. So you can imagine so in this king cloud service as an ecosystem of the chain. Node and task is trader and the workers.

B

So this is how we we hope to to give the whole network the computation capability, with this architecture.

B

Okay, I think that's it for in introduction to uh jobs tv today, if you have any questions, feel free to to listen. Let us know thanks.

A

Great thanks very much yeah um any questions for william of ken labs on what we've just seen.

A

Cool well yeah keep asking questions in the zoom chat or the slack channel if you've got any um and now let's hand over to nicola from crypto netlab to speak about what they've been working on.

C

Hello, just give me one second, that I'm going to share my scripts.

C

All right, so this is a public page, so you can get to see what's going on in here and so for those of you who know us cryptonet is a crypto applied, crypto research group at protoglobs and we're looking into uh traditionally we'll be looking into proof of space and file coins, protocol design, several authors of several faps and now we're looking into some ways to look at retrieval markets and we're focusing mostly on retrievability, so guaranteeing retroability and giving metrics on good retrieval nodes. So the project here is a storage metric style.

C

I don't know why we just decided to call it it's more of a consortium, but there are plans to make it more decentralized later on.

C

The idea is that if we can scan the network for uh for metrics- and we had a way to report this metrics in a way that could be aggregateable and unbiased, um then we would have on-chain metrics for miners, not just miners storage, not just storage providers, also ipfs modes.

C

So the thesis here is that if we can have on-chain metrics for storage providers, file coin or not, and these metrics could be realistic and objective, then we can add, rewards based on this metrics, not just us. Anyone can say we're going to give rewards to the top 10 miners, sorry to top 10 providers and and the goal is that we believe that the network will optimize towards this matrix.

C

So if we pick good metrics and we have a good way of reporting them, then we believe that this is going to be overall great for the network, because this is some sort of optimization function that everyone will will have the protocol in high level. It's uh it's. uh What we want to do, it's very simple which is: uh let's have a network of auditors, this network of auditors, they scan the network or they simply use the network.

C

Let me give you an example: s3 is a node in the network and they use the network constantly. They talk to a lot of storage providers and so on. So if you are a storage, if you are a user of the network that uses the network very intensively, then you should be an auditor in this network.

C

Then the idea is that there is an incentive mechanism for auditors to report through metrics. There are several ways that we're looking to uh some have to do with survey-based reporting some have to do with uh agreeing with majority, but in other words we are coming we're coming up with schemes where it is a an incentive not to collude through other nodes to other auditors in the network and the least you could lose the more rewards you get think more of like.

C

If you know where the a mine of gold is, you don't go and tell to everyone, and the idea is that if you know the good information about the network, you don't go. Tell anyone until you get rewarded for this.

C

So we want every player to be.

C

To build a lot of like network intelligence and share the results widely, but uh in a way that, if they earn rewards at the moment in which they share- and they don't want to share them earlier, what we do we aggregate this metrics on chain and then finally, we give incentives to storage providers that provide a good service.

C

A

C

The way that one, the way that we're looking into this is that there are five steps and we're like barely a step one and the five steps are the following at the very beginning. We want to have an open call for um partners that are already scanning the firecall network.

C

uh This is because we don't plan to build any software for scanning the network at least today, and these partners are, for example, the retrieval bot estery, maybe starboard or mile. Whoever is already built a lot of this infrastructure for scanning the network.

C

The goal is to have an aggregator node that gets all of these all of these signed statements, and then they aggregate the sign statements and they calculate a reputation score for each storage provider.

C

This step, I think it's just a bootstrap step. There are more details here. These are the groups that we're looking into talking to. If you believe that you're one of those storage uh auditor like reach out nikola protocol door, ai or just through this lack and uh yeah, so we're trying to agree on matrix and agreeing on how do we run the aggregator node and then just so that you understand the next steps?

C

The next steps is that the very beginning is going to be an initial consortium and we're only going to scan the file current network step two is to scan anyone, not just storage providers and file coin, but any ipfs node don't want to subscribe to being scanned. So what you do as a node, you say hey, I want to be scanned. These are the files that I claim that I have and then the network will start scanning you as well.

C

Then we have a third step, which is incentives for zero providers, which is the thing uh that once we calculate the score, we would like to get protocol labs and maybe the foundation and other partners to give some rewards to those that are doing uh great with their with their stats.

C

And the goal, of course, is to find the sponsors for the first round.

C

So once we have this incentivized the clear provider nodes, then we want to run the last step, which is incentivize auditors. At the very beginning, the auditors are just like an initial consortium and step four. We want anyone to become an auditor.

C

um Auditors should also be incentivized. I think that the way we're going to incentivize them is by having themselves a reputation, score and highly likely is non-tradable and, and you build uh the more reputation you build, the more you can participate with power in the network and then finally, this will go whenever fvm will launch. This should be a contract on fm and we will- and this will be up and running live.

C

uh There are some uh more documents in this doc. If you want to go and look at what are the metrics that we're looking into? What is our research? Where is the high level protocol spec, but that's it for the storage matrix.

A

Awesome thanks nichola um has anyone got any questions for nicola.

A

No all right! Well, thanks very much for joining today, guys, um I will put the recording up on youtube as soon as I can, and then we can share it with others who weren't able to make this this time slot thanks for joining and I'll catch, you guys later.