DCompute Off-chain Computing Outsourcing: Is EVM Enough?

1 Introduction

The current Ethereum L2 expansion solutions mainly include:

• Optimistic Rollup
• zkRollup: such as Polygon zkEVM, Scroll, etc.

These L2 schemes are still based on EVM.

However, high-performance applications such as games, on-chain order books, Web3 social networking, machine learning, and genome modeling are compute-heavy, which is still expensive for L2 EVM—EVM calculation is not fast enough, and its performance is not as good as Sealevel Virtual Machine.

If EVM is used in L3, the computing economy may be higher, but due to the structure of EVM itself, it cannot do parallel computing, so it is not the best choice for heavy computing. In keeping with the spirit of decentralization, new infrastructure (a new network of nodes) needs to be built before a new layer is built, which still requires the same set of providers to scale, or a whole new set of node providers (individual /enterprise) to provide resources, or both.

Therefore, every time a more advanced solution is built, the existing infrastructure should be upgraded, or a new layer must be built on top of it. To solve this problem, we need a post-quantum safe, decentralized, trustless, high-performance computing infrastructure that can use quantum algorithms to perform complex calculations in the real economy for decentralized applications.

Alt-L1s such as Solana, Sui, and Aptos can achieve parallel execution, but market sentiment, tight liquidity, and lack of developers in the market may not be called EVM killers due to the lack of a large trust moat built across Ethereum. As of now, the ETH/EVM killer may not exist yet.

The key question is: why do all calculations have to be on-chain? Is there a non-blockchain trustless, decentralized enforcement system? The answer lies in - the DCompute system can.

The DCompute infrastructure is decentralized, post-quantum secure and trustless, does not need and should not be blockchain/DLT technology, but it is very important to verify the calculation and generate the correct state transition and final result. The EVM chain will run exactly as it should, maintaining network security and immutability, while decentralized, trustless, secure computing can be off-chain.

This article does not focus on the issue of data availability. Because solutions like Celestia and EigenDA are already moving in this direction.

There are two main ideas for off-chain computing outsourcing:

• 1) Only outsourced computing: such as the current zkRollups, this solution is adopted, and data availability is still stored on the chain. The flaws are:

  • Either limited by the EVM
  • Either need to educate developers to learn a new language or instruction set

  The ideal solution should be high performance, cost and resource efficient, decentralized, private, and verifiable. It is usually to run calculations and build ZK proofs on cloud service provider AWS servers, but AWS servers are not decentralized. To this end, it is necessary to build a decentralized verifiable cloud computing network, such as: Nexus , Nillion . But it cannot be verified without ZK proofs.
  

• 2) While outsourcing computing, data availability is also outsourced: the off-chain computing and data availability layers are separated, but the computing still needs to be verified on the chain.
  

The current outsourcing computing solutions of the Ethereum ecosystem include:

In addition to the L2 expansion solution, there are:

• 1) Secure Enclave Computations/ Trusted Execution Environments (TEE): TEE is like a special box in a computer or mobile phone, which can only be accessed by specific trusted applications. Trusted applications running inside the TEE cannot be modified by other programs, or even the operating system itself.
• 2) Secure Multi-Party Computation (SMPC): The current MPC provider Sharemind can provide MPC infrastructure for computing, but the provider is still centralized.
  The basic process of SMPC is:
  • 2.1) Convert computing input into shares and distribute to SMPC nodes.
  • 2.2) The actual calculation is: interactive messages between SMPC nodes. Ultimately, each node has a share of the computed result output.
  • 2.3) These result output shares will be sent to one or more Result nodes - run LSS to reconstruct the final result value.
• 3) Nil Message Compute (NMC): A new distributed computing method developed by the Nillion team. As an upgraded version of MPC, nodes in NMC do not need to interact with each other's output. NMC uses a cryptographic primitive called One-Time Masking (OTM) to blind factors to mask a secret, similar to one-time padding.
• 4) ZK Verifiable Computation: The key to the Ethereum zkRollup expansion solution.
  

The comparison of different computing outsourcing solutions is as follows:

The key to ZK Verifiable Computation is:

• 1) Choose some kind of special proof primitive: cheap to generate, doesn't require a lot of memory, uncomplicated to verify.
  

• 2) Design a specific zk circuit based on the selected proof primitive.

• 3) Computational systems/networks: Given an input, run a specific function to generate a specific output.

Existing circuit programming languages ​​such as Circom and Cairo still have entry barriers for Web2 developers. The options for writing zk circuits using the existing language stack of Web2 developers are:

• 1) Lurk Labs : Adopt Nova certification system, compatible with Javascript developers. [Nova proof size will increase linearly with computation size, but the final proof can be further compressed by SNARK. 】
• 2) RiscZero : The STARK proof system is adopted, and the Bonsai network is provided. The proof can be verified on the chain: [STARK proof size does not increase with the computation size. 】
  
• 3) Nil Foundation : Create a decentralized verifiable outsourced computing infrastructure.

References

[1] Siddharth Rao’s June 2023 blog Is the EVM Alone Enough?
[2] Jacob Eberhardt et al.’s 2018 paper Off-chaining Models and Approaches to Off-chain Computations