atom_crypto

Posted on Oct 31, 2022Read on Mirror.xyz

Eclipse - Bring Solana VM to Next Modular Stack

This is Chloe, founder of CFG Labs. We are building the best builder community for the modular world. Myself developers, content creator, but most of the time, I have been an angel investor since 2016 after leaving hsbc investment banking team. Mathematics economics degree from UCL and Cambrdige. Today, we are honored to invite Neel Somani, CEO of Eclipse, one of the most innovative protocols bringing solana vm to the modular space, just secured the big money from top investors like Polychain, solana foundation etc.

I think we have seen lots of innovation in infrastructure back to 2016. various layers trying to improve user experience either through throughput, finality, latency, scalability in sacrifice of decentralization and security , etc. So we believe we are now transitional into the next huge opportunities.

  1. So before we get deep into that. Could you make a quick intro of yourself, what eclipse building, and what’s the progress

I am Neel Somani, my background is in traditional finance. I work in Citadel as a commodity group.worked in power and gas group.That’s entirely non-crypto staff. look at different market structures, modeling out what power, gas price being. But that’s very relevant to crypto now.people do frequent batch auctions, all kinds of different market structures. I quit in March, I previously did in Terra ecosystem. I was building an EVM, that’s the way to run EVM code on Terra.Terra debugging. I talked to Celestia, and started to buy Celestia's vision of the future.that's a far more scalable and decentralized future than the direction I am heading right now with the cycles of L1. I was in Chicago at the time, chatting with the Celestia core team. I was talking with the team heading into the scalability issues on the solana blockchain itself, Despite the fact, solana being one of the fastest blockchain out there.they were using 10% of the network, for example,pyth the oracle of solana, recently spin out the first solana app chain called pythnet, run the infrastructure themselves they have to maintain the reliability. .I think they are really regretting that because of the application layers, they actually shouldn't have been thinking about that. That's the problem eclipse is solving. Application needs more throughput. They need to have more complete control of the stack. That's what we offer, the customizable modular rollup. They use the solana vm. So you get all the throughput that provides, you also get to choose where the security comes from. So we started from Celestia, because Celestia is the best DA layer, we support Solana, we support a bunch of other DA layers as well.

Chloe

That’s great, to be honest we don't have too much open information right now, but like the mix & match modularity design, sovereign community, minority nodes assumption under the trust minimized environment, are something very impressive.

  1. I want to talk more about Solana. Solana is something we are actually interested in, they have supported great DEX projects, NFT projects, like Magic Eden. What do you think of its huge potential in terms of engineering design, ecosystem development/use case etc.

Solana vm is the best execution engine machine, that’s why we chose them. In fact, the earliest version of this project involves building our own dspoke virtual machine. We are basically designing from scratch. That looks very similar to Solana. anatoly yakovenko, Solana foundation gives us grants. Given the incredible support from the Solana ecosystem, it doesn't make sense for us to build the virtual machine from scratch. That was the thinking there. There was the great fee market tuning up, doing parallelization across transactions, using GPU for signature verifications,all the fancy stuff solana execution layer does. But the tricky thing is that Solana isn't built to be a rollup. the repo model doesn't actually factor out the consensus from execution. Other parts are there are no light nodes for solana, where you can normally do for very cheap verification of solana transactions. That’s not the thing. So we have to make sustainable changes to the Solana code base, in order for it to work, without infrastructure.

CFG Labs The scalability of Solana, relies on 1) efficient use of the bandwidth 2)reduce the communciatoon overhead between validators, and 3) improve the execution speed. POH is interesting because it provides the immutable, censorship resilience global ording, for different messages and input, which is used to improve the propagation experience in Turbine block propagation protocol. But there are still some problems to be solved, 1) The frequent collapse of Solana lead nodes 2) The MEV attack in Solana

For the use cases. I am actually big fans of socials, games etc. We have seen many iterations of gaming for the past few years. By experience, The play to earn model is cool but not long term sustainable, but we can see new projects keep iterating ,and innovating in the space. We believe over 55% of the global players are in Asia pacifics and the region has strong potential to gain the stronger market share in the future. For the NFTs, they really enable the digital tokenization of assets like art, media, in-game items, identity, financial products, and physical/digital real estate. Their ownership is secure and provable on the blockchain bringing an entirely new mechanism for digital rights management. I think there are some good projects, such as the naming service DID project built for the modular world, that will bring real digital identity and data privacy to the sovereignty community.

3)I just Solana is great, but there are some problems 1) The frequent collapse of Solana lead nodes 2) The MEV attack what do you think the potential solutions for that

Neel: For the optimization, mix and matching, you mention turbine block production, we don't use that, given there is only 1 leader, we are starting with the single sequencer. so there is no one to propagate the block to. if that makes sense. The Sequencer produces the block, and stores it entirely on the celestia. So we don't use Turbine. Proof of history, the purpose of Solana is actually decoupling execution from consensus. like the well structured similar way our rollup do. The last point is Sealevel we use. When you were mentioning the issues of Solana, I guess Gulf Stream, they removed the mempool and they forwarded the transaction to the leader. We actually have the mempool. Some of the optimization we dont need for Eclipse,The real unlock is based on the fact that you can horizontally scale and each application gets execution. I also think the fee market, they just recently switch to …, using UDP, Solana is always trying to solve the liveness situations on L1. but we actually come into the long tail applications. that cant be run on solan L1. even if they run correctly. That;s thes staff like you are mentioning the NFT on Solana, typically because Solana has low fees. but if as the NFT product, if you really want to optimize for it, you need to modify it for the sequencer. for example. you might slow down the block time for a NFT drop because you don't want bots pick up NFTs because it a little bit faster. or you can apply some batch commitment scheme that when there are bunch of people trying to buy NFT at the same time, you have some consequence failures. you fail the orders in some reasonable ways. On the solana L1, you can’t get the features anyway. Yeah, we are borrowing a lot of NFT infrastructure and several projects are coming to Eclipse.

So for lead collapse, the problem is actually the block is produced very fast, the lead can produce blocks one by one ,without completing the whole consensus process. So the follower node can’t catch the speed. The average block production time in Solana is 1s while in Ethereum 12s. So the problem can be alleviated by increasing the block production time. I think Sui,Aptos are trying to solve the problem.

For the Solana, we believe MEV is also a concern as it further intensifies the lead validators problem. The reason is that 1) there is no clear gas fee mechanism as those in Ethereum, so searchers have to submit repetitive transactions to the lead node, in hope of being added during the NFT mining or huge front run opportunities. 2) Under the mempool-less transaction design, transactions can’t be filled efficiently as those bloom filters in Ethereum, also based on the fact the lead node for the next slot is chosen ahead and publicly known in advance.

  1. What kind of technology you could share publicly to facilitate the trust-minimized app chain and multi chain environment

The first part is its different types of rollups. They describe rollups as modular, but they don't understand what it even means.The big unlock is that we don't do settlement on-chain. We have our own settlement which is built on Solana VM. That’s the one feature. The second feature is we have zero knowledge solana vm, we are the first one to bring that to the market. We will also enable IBC for the solana VM, that’s the condition for the Solana foundation grant. Those are the kind of tech we are building right now. Apart from implementing the settlement for the Solana VM, you have to work in the parallel environment the solana operats. different from EVM, which is singe thread. that’s a lot of simpler to implement optimistic settlement for evm chains.

CFG Labs: dYmension is interesting, as it introduces IRC, IBC based communication protocol within their own environment. In order to service multiple rollup implementations, dYmension settlement layer spins up a virtual machine containing the RollApp execution environment upon fraud dispute. For example, if there are rollups that support the Solana VM execution environment, you don’t have to permanently add SolanaVM into the settlement layers. Post RVM emulation the container is discarded. RVM assures that any rollups built on top that may always evolve, experiment and keep up to date with state of the art advancements in blockchain technology and execution in particular.

There is the light paper I am writing with Mustafa, Celestia team back in June, there are some pretty of these projects. dYmension and He, evm settlement layers.

  1. MEV and value accrual is a very interesting topic, before we dive more into that, can we give more details about how the eclipse accrual its value?

The way we think of the value accrual in the modular world first contrasting how we work in the monolithic world where there is one protocol, and every application lives there. Any if L1 really judges the fees, wasn't anywhere for applications to go because that’s the only choice we want to pick the decentralized or particular execution layer. Now you have some DA layer, but you have multiple execution layer, then you switch the DA layer very easily. it;s not like value accrual with DA layer, but not like perfect competition. da layer needs to compete on some other attributes, the level of decentralization, the sophistication of DA sampling. there might be DA solutions completely centralized. For example, you can imagine onchain Twitter may provide the DA solutions, they may charge no money basically. they can competing on price.The argument is that no one reads the smart contract anyway when they interact with the protocol, you implicitly trust with the twitter data for your blocks as well. DA come from a variety of solutions massively decetnalizated, very expensive, vs total centralized but very cheap. then you have to think of settlement and execution.where the value accrual between those two. settlement is based on appfex., more chains on the settlement layers, the more valuable the settlement layer becomes. you can move money between the chains. execution depends on the mode execution layer provides. starkware closed source, i don't know if you review that. They already have sophisticated zk proof techniques. they have bespoke vm cairo,once you write for Cairo, you cant move anywhere else. as the result of they charge more for the execution layer, you have to pay for it. that area kinds of execution layer accrual value on the long run. something like technical edge, stikckess on y, if you build one application layer on one chains, you cant just put on another.

so maybe a super fast vm execution layer could accrue value because there is a game use all the throughput, they can just move to Ethereum but ethereum don't support the level of throughput you need, and in that case you can charge more.that's the way execution layer can gain pricing power.

That’s true. I think shared security is one thing I would love to mention. Security I am referring to here is about DA security and Consensus Security. For example, Celestia works as the data shared security layer, which is similar to the Cosmos Hub, interchain security where it works for consensus and validators service and is getting paid by the consumer chain. So I believe the celestia DA nodes should be incentivized property. Right now, only storage nodes are incentivising for storing data, I believe every rollup, settlement layer built on top of that should contribute a% of fee to the celestia and a more well-structured economics design should be introduced by the community. Also, the introduction of celestia allows the segregating of the different markets under different fee markets, which improves the efficient resource pricing problem we normally see in the monolithic world. Other innovative projects too look at such as eigen layers A platform to leverage Ethereum security through the innovative method of restaking.

And for Some rollups infrastructure like Eclipse and others, I believe you mentioned you are building either the settlement layer, as the hub for rollups, but also the rollup itself. We believe there are strong value accruals as more dapp developers are building on top of that. The transactions are paid by the users, also there is a fee for IBC similar interoperability between rollups, sequencers have to submit data block and state roots, also sequences have to stake % weight on the network.

That’s all correct, you charge for all the business. We provide execution, settlement, you charge, for that, expenses of having the data post of the data availability layer, but you can get additional revenue right. We are working with skip,they are able to capture the MEV and we are able to turn it back to our execution layer. That app specific execution layer benefits from the MEV being produced.

  1. Future of value accrue for Modular economy.

MEV is completely sophitictc in the monolithic chain where all applications have to get together just because of more opportunities. you only have one application on the chain.that hard to say I should look at more . Straight transactions volumen people so sophisticated transactions volume people are trading against the chain, but necessary setting the slippage to the exact value they are comfortable with. MEV is an opportunity for us because we can capture the MEV as the protocol and we turn it either to eclipse protocol or those building on us. we can probably capture the protocol. It’s not some primary we are concerned with.

At the Ethereum developers conferenceTarun from Gauntlet gave a very interesting topic for the theory of MEV. How to balance the equilibrium of users benefits from the social welfare point of view, and for validators from the economic security point of view. Right now there are different kinds of MEV, sandwiches,liquidations, artbitragues, NFT mint front running, cross chain. But, it's hard to say whether it's good or not. For example, there is some research proving sandwiches can also improve social welfare where some Liquidity in time improves the user experience, but sacrifices the benefits for the passive liquidity providers.

Some projects such as the Skip protocol are very interesting. They are bringing cross-chain MEV solutions to the Cosmos. Some of the existing problems for PBS in Ethereum, is that block builders in PBS will include toxic forms of MEV will win out, which will harm the user experience. They are also working on-chain flashbot version solutions. I am really excited about that.

I think time is there. It's a very nice talk. I learn a lot. Thanks again for Neel, and all the best to the Eclipse future. and I look forward to talking next time

Solana