The future of L2 and ALEO zero-knowledge proof (ZKP) mining

Source: Wu Says Blockchain

In the public chain track, the route of zero-knowledge proof (ZKP) has become very clear. In the future, it is highly likely that a situation will form where Ethereum L1 is at the core to ensure security, while ZKP is deployed on L2 to achieve privacy functions. Public chains using ZKP besides L2 include Aleo, Mina, and Filecoin.

As Ethereum ends POW, and ZKP requires a large amount of computing power to maintain, it is foreseeable that in the near future, ZKP will take on a large amount of GPU computing power. Messari analyzed the feasibility of ZKP taking over Ethereum's computing power in the June article "What Will Ethereum Miners do After The Merge?"

"As users seek more expressive, high-performance, and private computing, the complexity of using ZKP will increase. This will lead to a slowdown in the speed of generating proofs, requiring specialized hardware to generate proofs in a timely manner. When Bitcoin was first launched, anyone with a standard CPU/GPU could mine Bitcoin. Eventually, professional miners developed more efficient hardware (ASIC), making CPU/GPU mining unprofitable. ZK mining is likely to follow a similar path, starting with standard GPU miners and then developing more efficient mining machines (ASIC or FPGA). ZKP is still in its infancy, but Paradigm predicts that the ZK miner/prover market could grow to a scale comparable to the PoW mining market in the future."

Compared to FPGA and GPU, the hardware cost of top FPGA is only 1/3 of that of top GPU (leading process nodes, clock frequency, energy consumption ratio, and memory bandwidth), and the energy consumption ratio exceeds that of GPU by 10 times. In comparison to FPGA and ASIC, ASIC is one-time programmable, while the code of ZKP is not as stable and unchanging as Bitcoin, and there are differences in the logic of different ZKPs, making ASIC obviously unsuitable for this variable business. FPGA, on the other hand, can modify the program simply by refreshing, which means it can be compatible with multiple sets of ZKP logic.

Therefore, in the longer term, FPGA may dominate ZKP mining; however, FPGA programming is quite difficult, and the technology is not yet mature. In the short term, GPU will be the main equipment for ZKP mining. For those former Ethereum miners, they naturally have a greater commercial incentive to support ZKP public chains rather than investing their computing power into abandoned public chains like ETC or ETHW.

L2 ZKP

Currently, the Proof and Batch submission process of ZK-Rollup is centralized, with a single Sequencer deciding the order of submissions. To scale the number of miners, this process needs to be decentralized, meaning any miner can submit Proof to L1. Regarding how to achieve decentralization, Vitalik has proposed some suggestions, such as establishing a DAO to conduct regular Sequencer decision power auctions, or randomly assigning it to one of the nodes with the probability of being selected proportional to the amount of ETH it deposits into the Rollup. However, regardless of the method, it essentially belongs to the POS mechanism, where income is determined by the size of the stake, which is different from traditional mining methods (where income is determined by the size of computing power).

Currently, the main L2s using ZK-Rollup include zkSync and StarkNet.

The daily operation of zkSync depends on the computing service provider for generating zero-knowledge proofs for block generation, namely "Stichting ZK Sync," a non-profit Dutch foundation registered in Amsterdam. According to the development roadmap of the team Matter Labs, zkSync plans to achieve decentralization in the future by introducing an independent consensus mechanism with two different roles: validators and Guardians. Validators are responsible for packaging transactions into blocks and generating zero-knowledge proofs for them, and their nodes must operate in a secure environment with good internet bandwidth, which is what we usually understand as large mining pools.

Guardians are token holders of zkSync, nominating validators based on token shares. Guardians' nodes can run on ordinary computers or cloud servers without the need for specialized service providers, making them less susceptible to censorship. Conversely, Guardians can select uncensored validator nodes by monitoring transaction data.

StarkNet also needs to issue tokens on its path to decentralization. Currently, StarkNet transaction fees are paid in ETH, and in the future, it may use StarkNet's native token, possibly supporting both native token and ETH payments. Besides being used to pay transaction fees, tokens can also be staked to make main voting decisions on transaction ordering, STARK proofs, etc., enhancing network integrity and censorship resistance.

Referring to a proposal made by Ohad: adding a BFT-based sequencer, where the sequencer is selected based on their shares, choosing a specific sequencer for each slot. During this process, the network will add a slashing rule (similar to Cosmos SDK) to couple incentives between sequencers and provers in the best possible way.

ALEO ZKP

Another popular project in ZKP mining this year is ALEO. ALEO is a privacy public chain, and currently, no public chain in the privacy public chain track has stood out. Whether it is Oasis, Secret Network, or Platon, none have truly realized the privacy functionality of dapps. ALEO is no exception; it currently has no practical value. However, unlike other privacy public chains, its consensus mechanism is similar to traditional PoW, requiring computing power to maintain. The difference is that the underlying computation for ALEO mining is not an arbitrary hash function but a proof of knowledge, so it can serve as PoW to ensure network consensus while also providing verification of transactions included in a given block.

ALEO has a significant advantage over L2 ZKP in that its mining algorithm is simpler. When verifying a block, the algorithm generates a random number, and if the final result after ZK computation is less than the target value, the computation is correct. The node that completes the computation first has the right to produce the block. Each time a mining machine verifies a block, it only needs to generate a random number to start the computation, and this mining logic is not much different in form from Bitcoin.

However, L2 ZKP computation is different. L2 faces a batch of transactions, so it needs to perform ZKP for each transaction during verification and then package and upload to L1. This process involves the issue of parallel computation, as even a single powerful CPU cannot verify thousands of transactions one by one; it must use devices like GPUs that have multiple computing units for parallel computation. However, although GPUs are currently the most suitable devices for ZKP mining, further algorithm optimization is needed to achieve parallel computation for L2 mining. ALEO has a significant advantage in this regard, as it does not require parallel computation, allowing GPU mining machines to integrate almost seamlessly. In today's immature privacy public chains and ZK L2, those idle ETH mining machines may have more motivation to connect their computing power to ALEO mining pools.

However, all of this is based on the premise that ZK L2 is not yet mature. Due to the prosperity of the ETH ecosystem, as long as L2 goes live, dapps on L1 can be directly migrated over. ALEO, on the other hand, even if privacy public chain technology matures, the team still needs to invest a large amount of money to build the entire ecosystem, which is obviously a significant disadvantage for ALEO. We can refer to this as the ecological moat of the ETH system: regardless of the type of L2, as long as it can add ZK on top of the original underlying public chain technology, there is a natural value capture ability. This is something that L1s like ALEO do not possess.

Conclusion

As the demand for censorship resistance gradually expands, the future privacy track is bound to become a revolutionary application. Currently, the most feasible in privacy proofs is ZKP, which requires a large amount of computing power to maintain, so ZKP mining is likely to become an industry comparable in scale to the initial ETH mining.

The biggest issue that ZKP mining needs to solve is decentralization, and currently, zkSync and StarkNet have released relevant roadmaps. Logically, to achieve decentralization, issuing tokens is essential. Tokens are not only used to maintain consensus but can also be used to supervise mining nodes, thereby increasing censorship resistance to some extent.

Among mining devices, the most cost-effective is FPGA; however, considering the high difficulty of FPGA programming and the large number of idle GPU mining machines, in the short term, GPUs are more likely to dominate this industry.

In public chains, ZKP mining can be adopted by any ZK series L2, but currently, only ALEO allows it at L1. In contrast, ALEO does not require parallel computation, so idle GPU mining machines can integrate seamlessly. L2, on the other hand, has more advantages in terms of ecosystem, with a larger user base and lower risks.

References:

https://www.paradigm.xyz/2022/04/zk-hardware

https://messari.io/report/what-will-ethereum-miners-do-after-the-merge

https://docs.zksync.io/userdocs/decentralization/

https://community.starknet.io/t/starknet-decentralization-tendermint-based-suggestion/998