Summarizing nine parallel EVM studies: What are the focus of discussion?

Written by: 0xNatalie

TL;DR

1. Parallel EVM has attracted the attention of top venture capital firms, and many projects have begun to explore this direction, such as Monad, MegaETH, Artela, BNB, Sei Labs, Polygon, etc.

2. Parallel EVM is not only about parallel processing, but also about comprehensive performance optimization of each component of EVM. Through these optimizations, more complex and efficientBlockchainApplication support.

3. Parallel EVM needs to achieve differentiation in the open source ecosystem, balance the needs of decentralization and high performance, and solve potentialSafetyDue to the complexity of multi-threaded programming, parallel EVM faces the difficulty of coordinating multiple transactions at the same time, so an effective solution must be found to ensure the stability andSafetysex.

4. In the future, parallel EVM will promote the implementation of on-chain central limit order book (CLOB) and programmable central limit order book (pCLOB), which will greatly enhance the efficiency of DeFi activities, and the DeFi ecosystem is expected to grow significantly.

5. By introducing alternative high-performance virtual machines (AltVM) into the Ethereum ecosystem, performance andSafetyThe performance will be significantly improved. This integration can make full use of the advantages of each virtual machine and promote the further development of Ethereum.

This year, Parallel EVM has attracted bets from top venture capital firms such as Paradigm and Dragonfly, and has attracted widespread attention from the market. Parallel EVM is an optimized upgrade of traditional EVM. Traditional EVM is serially processed, that is, transactions must be executed one by one in sequence. This method is prone to transaction congestion and processing delays during peak trading periods, affecting user experience. Parallel EVM introduces parallel processing technology, allowing multiple independent transactions to be executed simultaneously, thereby greatly improving transaction processing speed. With the development of full-chain games and account abstraction,walletThe emergence of complex applications such asBlockchainPerformance requirements continue to increase. In order to support a wider range of users,BlockchainThe network must be able to process a large number of transactions quickly and efficiently. Therefore, the parallel EVM is essential as a solution for the growing Web3 applications.

However, implementing parallel EVM faces the following common challenges that require precise technical solutions to ensure stable operation of the system:

• Data consistency: In parallel EVM, because multiple transactions occur at the same time, multiple operations may need to read or modify account information at the same time. In order to ensure that each operation is correctly reflected in the final state, an effective locking mechanism or transaction processing mechanism is required to ensure data consistency when the state is modified.

• State access efficiency: Parallel EVM needs to be able to quickly access and update state, which requires the system to have an efficient state storage and retrieval mechanism. By optimizing the storage structure and access path, such as using more advanced data indexing technology and caching strategies, the latency of data access can be significantly reduced and the overall system performance can be improved.

• Transaction conflict detection: In parallel execution, multiple transactions may depend on the same data state, which makes the order and dependency management of transactions complicated. To this end, complex scheduling algorithms must be designed to identify and manage dependencies between parallel transactions, detect potential conflicts, and decide how to handle them to ensure that the results of parallel transactions are consistent with those of serial execution.

For example, MegaETH decouples transaction execution tasks from full nodes and assigns different tasks to specialized nodes to optimize overall system performance; Artela uses predictive optimistic execution and asynchronous preloading technology to AI Analyze the dependencies between transactions and load the required transaction states into memory in advance to improve state access efficiency; BNB Chain has developed a dedicated conflict detector and re-execution mechanism to improve the system's ability to manage transaction dependencies, reduce unnecessary re-execution, etc.

To gain a deeper understanding of the development direction of parallel EVM, the following nine high-quality articles on this topic are selected, providing a comprehensive perspective on specific implementation plans for different chains, comprehensive ecological research, future prospects, etc.

MegaETH: Unveiling the First Real-Time Blockchain

Author: MegaETH, Date: 2024.6.27

MegaETH is an EVM-compatible Layer 2 that aims to achieve near-real-time performance of Web2 servers. Its goal is to drive Ethereum L2 MegaETH provides high transaction throughput, ample computing power, and millisecond response times, allowing developers to build and combine complex applications without performance limitations.

MegaETH improves performance by separating transaction execution tasks from full nodes and introducing parallel processing technology. Its architecture consists of three main roles: sequencer, validator, and full node.

• Sequencer: The responsibility of the sequencer is to sort and execute transactions submitted by users. After executing the transaction, the sequencer will send the transaction status change (status difference) to the full node through the peer-to-peer (p2p) network.

• 全节点（Full Node）： 全节点接收到状态差异后，直接应用这些差异来更新自己的本地区块链状态，而不需要重新执行交易。这大大减少了计算资源的消耗，提高了系统的整体效率。

• Prover: Prover verifies blocks using a stateless verification scheme. Instead of verifying blocks one by one in sequence, multiple blocks can be verified at the same time. This method further improves verification efficiency and speed.

This node specialization design allows different types of nodes to set independent hardware requirements according to their respective functional requirements. For example, sequencers require high-performance servers to process a large number of transactions, while full nodes and validators can use relatively low-configuration hardware.

Presenting Artela Scalability Whitepaper – Parallel Execution Stack and Elastic Block Space

Author: Artela, Date: 2024.6.20

Artela has significantly improved the parallel execution efficiency and overall performance of blockchain by introducing a number of key technologies:

• Parallel execution: By predicting transaction dependencies and grouping transactions, multiple CPU cores can be used for parallel processing to improve computing efficiency.

• Parallel storage: Optimizes the storage layer, supports parallel data processing, avoids storage bottlenecks, and improves overall system performance.

• Elastic computing: Supports multiple computers working together to achieve elastic computing nodes and block space, providing dApps with higher transaction throughput and predictable performance.

Specifically, Artela’s predictive optimistic execution leverages AI Intelligent analysis of transactions andcontractThe system dynamically accumulates and stores the state access information of historical transactions for use by the prediction algorithm. Through asynchronous preloading, the system loads the required transaction state into memory in advance to avoid I/O bottlenecks during execution. At the same time, parallel storage optimizes Merkelization and I/O performance by separating state commitment from storage operations. This separation processing enables parallel and non-parallel operations to be managed independently, further improving the parallelization efficiency of the system.

In addition, Artela built Elastic Blockspace (EBS) based on elastic computing. In traditional blockchains, all dApps share the same blockspace, resulting in resource competition between high-traffic dApps, unstable gas fees and unpredictable performance. Elastic Blockspace provides dApps with exclusive and dynamically scalable blockspace to ensure predictable performance. dApps can apply for exclusive blockspace according to demand. As the blockspace increases, validators can expand processing power by adding elastic execution nodes to ensure efficient resource utilization and adapt to different transaction volumes.

Road to High Performance: Parallel EVM for BNB Chain

Author: BNB Chain, Date: 2024.2.16

On the BNB chain, the team has taken multiple steps to implement parallel EVM to improve transaction processing capabilities and scalability. The following are the main tasks done by the BNB chain in the process of implementing parallel EVM:

Parallel EVM v1.0

In early 2022, the BNB chainCommunityIt promotes the execution of parallel EVM. The core components include:

• Scheduler: Responsible for allocating transactions to different threads for parallel execution, thereby optimizing throughput.

• Parallel Execution Engine: Leverages parallel processing to execute transactions independently on dedicated threads, significantly reducing processing time.

• Local state database: Each thread (the basic unit of independent execution of instructions) maintains its own "thread-local" state database to efficiently record state access information during execution.

• Conflict Detection and Re-Execution: Ensure data integrity by detecting and managing transaction dependencies and re-executing transactions when conflicts occur to ensure accurate results.

• State submission mechanism: Once the transaction execution is completed, the result will be seamlessly submitted to the global state database, updating the overall state of the blockchain.

Parallel EVM v2.0

Based on the parallel EVM 1.0, BNB chainCommunityA number of performance-enhancing innovations have been introduced:

• Streaming Pipeline: Improves execution efficiency and enables transactions to be processed smoothly in a parallel engine.

• Universal Unconfirmed State Access: By optimizing the access to state information, other transactions are allowed to temporarily use the results of a transaction before it is completed but not officially confirmed, thus reducing the waiting time between transactions.

• Conflict Detector 2.0: An enhanced conflict detection mechanism that improves performance and accuracy, ensuring data integrity while reducing unnecessary re-executions.

• Scheduler Enhancements: The scheduler now leverages static and dynamic scheduling policies to more efficiently distribute workloads and optimize resource utilization.

• Memory optimization: Through shared memory pool and lightweight copy technology, memory usage is significantly reduced, further improving system performance.

Parallel EVM v3.0

在并行 EVM 2.0 取得性能提升之后，BNB 链Community积极开发并行 EVM 3.0，目标如下：

• Reduce or eliminate re-executions: By introducing a hint-based scheduler, external hint providers are used to analyze transactions and generate predictions of potential state access conflicts. With these hints, transactions can be better scheduled with fewer conflicts, thereby reducing the need for re-executions.

• Modularity: Breaking the code into independent modules improves maintainability and facilitates adaptation to different environments.

• Codebase refactoring: Align the latest BSC/opBNB codebase to ensure compatibility and simplify integration.

• Thoroughly tested and validated: Extensive testing under various scenarios and workloads ensures the stability and reliability of the solution.

Sei's Parallel Stack

Author: Sei, Date: 2024.3.13

Sei Labs has created an open source framework called Parallel Stack, which aims to build Layer 2 solutions that support parallel processing technology. The core advantage of Parallel Stack lies in its parallel processing capabilities, which leverages the advancement of modern hardware to reduce transaction fees. The framework adopts a modular design, allowing developers to add or modify functional modules according to specific needs to adapt to different application scenarios and performance requirements. Parallel Stack can be seamlessly integrated with the existing Ethereum ecosystem. Applications and developers using Parallel Stack can directly use Ethereum's existing infrastructure and tools without doing a lot of modification or adaptation work.

To ensure transactions and intelligencecontractofSafetyParallel Stack introduces a variety of security protocols and verification mechanisms, including transaction signature verification, smart contract auditing, and abnormal transaction detection systems. To facilitate developers to build and deploy applications on Parallel Stack, Sei Labs provides a complete set of developer tools and APIs to support developers in making full use of the high performance and scalability of Parallel Stack and promote the further development of the Ethereum ecosystem.

Innovating the Main Chain: a Polygon PoS Study in Parallelization

Author: Polygon Labs, Date: 2022.12.1

The Polygon PoS chain doubled its transaction processing speed by introducing a parallel EVM upgrade, thanks to the metadata minimization method.

Polygon borrowed the principles of the Block-STM engine developed by Aptos Labs and developed a minimal metadata approach to meet Polygon's needs. The Block-STM engine is an innovative parallel execution mechanism that assumes that there are no conflicts between all transactions. During transaction execution, the Block-STM engine monitors the memory operations of each transaction, identifies and marks dependencies, and reschedules conflicting transactions for verification to ensure the correctness of the results.

The minimal metadata method records the dependencies of all transactions in a block and stores them in a directed acyclic graph (DAG). Block proposers and validators first execute transactions, record dependencies and attach them as metadata. When blocks are propagated to other nodes in the network, the dependency information is stored in the blockchain.Xiaobai NavigationThe information is already included, reducing the computational and I/O burden of re-verification and improving verification efficiency. Through the minimum metadata method, recording dependencies also optimizes the transaction execution path and minimizes conflicts.

What is the significance of EVM parallelization? Or is it the end of EVM hegemony?

Author: Zhixiong Pan, Founder of ChainFeeds, Date: 2024.3.28

Parallel EVM technology has attracted attention and investment from some top venture capitals including Paradigm, Jump and Dragonfly. These capitals are optimistic that Parallel EVM can break through the performance limitations of existing blockchain technology and achieve more efficient transaction processing and wider application possibilities.

Although the literal meaning of Parallel EVM is just "parallelization", it actually covers much more technical improvements than that. Parallel EVM not only allows multiple transactions or tasks to be processed at the same time, but also includes in-depth performance optimization of various components of Ethereum EVM. For example, improving data access speed, improving computing efficiency, optimizing state management, and many other aspects. Therefore, its efforts are likely to represent the performance limit under the EVM standard.

In addition to technical challenges, parallel EVM also faces the difficulties of ecosystem construction and market acceptance. It is necessary to create differentiation in the open source ecosystem and find the right balance between decentralization and high performance. In terms of market acceptance, it is necessary to prove to the industry and the outside world the performance improvement and cost-effectiveness brought by its parallelization capabilities, especially in the context of a large number of existing Ethereum applications and smart contracts that are already running stably, the incentive to switch to a new platform needs to be very prominent. In addition, the promotion of parallel EVM also needs to solve potential security issues and possible new technical defects to ensure the stability of the system and the security of user assets, which are important factors in promoting the widespread adoption of new technologies.

Death, Taxes, and EVM Parallelization

Author: Reforge Research, Date: 2024.4.1

With the introduction of the parallel EVM, the feasibility of on-chain Central Limit Order Books (CLOB) has been improved, and DeFi activity is expected to increase significantly.

In CLOB, orders are prioritized by price and time to ensure market fairness and transparency. However, when implementing CLOB on blockchain platforms such as Ethereum, it often leads to high latency and high transaction costs due to the limitations of platform processing power and speed. The introduction of parallel EVM greatly enhances the processing power and efficiency of the network, enabling DeFi trading platforms to achieve faster and more efficient order matching and execution. CLOB becomes feasible.

On this basis, the Programmable Central Limit Order Book (pCLOB) further expands the functionality of CLOB. pCLOB not only provides basic buy and sell order matching functions, but also allows developers to embed custom smart contract logic in the submission and execution of orders. These custom logics can be used for additional verification, determination of execution conditions, and dynamic adjustment of transaction fees. By embedding smart contracts in the order book, pCLOB provides greater flexibility and security, supporting more complex trading strategies and financial products. By leveraging the high performance and high parallel processing capabilities provided by the parallel EVM, pCLOB is able to implement complex and efficient trading functions similar to traditional financial trading platforms in a decentralized environment.

However, despite the significant improvement in blockchain performance brought by the parallel EVM, the security of the existing Ethereum Virtual Machine (EVM) and smart contracts is still insufficient and vulnerable to hacker attacks. To address these issues, the authors propose a dual VM architecture. In this architecture, in addition to the EVM, an independent virtual machine (such as CosmWasm) is introduced to monitor the execution of EVM smart contracts in real time. This independent virtual machine functions similarly to antivirus software in an operating system, capable of advanced detection and protection, thereby reducing the risk of hacker attacks. In the future, emerging solutions such as Arbitrum Stylus and Artela are considered promising to successfully implement this dual VM architecture. With this architecture, these new systems can better embed real-time protection and other critical security features from the beginning.

What will be the next step towards enhanced scalability while maintaining EVM compatibility?

Author: Grace Deng, researcher at SevenX Ventures, Date: 2024.4.5

New Layer 1s such as Solana and Sui offer higher performance than traditional Layer 2 and Layer 1 by using a new virtual machine (VM) and programming language, and adopting parallel execution, new consensus mechanisms, and database designs. However, these systems are not EVM-compatible, resulting in insufficient liquidity and higher barriers for users and developers. But EVM-compatible Layer 1 blockchains such as BNB and AVAX, despite improvements at the consensus level, have fewer modifications to the execution engine, so performance improvements are limited.

Parallel EVM can improve performance without sacrificing EVM compatibility. For example, Sei V2 improves read and write efficiency by adopting optimistic concurrency control (OCC) and introducing a new state tree (IAVL trie); Canto Cyclone optimizes the state management system by using the latest Cosmos SDK and ABCI 2.0 technology and the IAVL state tree in memory; and Monad proposes a new Layer 1 solution that combines high throughput, decentralization, and EVM compatibility, using OCC, a new parallel access database, and the MonadBFT consensus mechanism based on Hotstuff.

In addition, it is also possible to consider integrating other high-performance virtual machines (AltVM) into the Ethereum ecosystem, especially those that support Rust development, such as Solana's Sealevel or Near's WASM-based VM. This will not only solve the drawbacks of EVM incompatibility, but also introduce Rust developers into the Ethereum ecosystem, improve overall performance and security, and open up new technical possibilities.

Ten thousand words to interpret parallel EVM: How to break through the blockchain performance bottleneck?

Author: Gryphsis Academy, Date: 2024.4.5

Parallel EVM is mainly to optimize the performance of the execution layer. It is divided into Layer 1 solution and Layer 2 solution. The Layer 1 solution introduces a transaction parallel execution mechanism to allow transactions to be executed in parallel as much as possible in the virtual machine. The Layer 2 solution essentially uses the parallelized Layer 1 virtual machine to achieve a certain degree of off-chain execution + on-chain settlement. In the future, the Layer 1 track may form two major camps of parallel EVM and parallel non-EVM, and the Layer 2 track will develop in the direction of blockchain virtual machine simulator or modular blockchain.

Parallel execution mechanisms are mainly divided into the following three categories:

• Message passing model: Each actor can only access its own private data, and access other data through message passing.

• Shared memory model: Memory locks are used to control access to shared resources, which is divided into memory lock model and optimistic parallelization.

• Strict state access list: Based on the UTXO model, the account address to be accessed by each transaction is calculated in advance to form an access list.

Different projects have adopted different strategies when implementing parallel execution mechanisms:

• Sei v2: Switches from a memory lock model to an optimistic parallelization model to optimize potential data contention.

• Monad: Introduces superscalar pipelining technology and improved optimistic parallelism to achieve performance of up to 10,000 TPS.

• Canto: Introducing optimistic parallelization with the Cyclone EVM and innovating on decentralized financial infrastructure.

• Fuel: As a modular Ethereum rollup operating system, it adopts the UTXO model and optimistic parallelization mechanism to improve transaction throughput.

• Neon, Eclipse, and Lumio: Provide cross-ecosystem performance improvements by integrating the capabilities of multiple Layer 1 chains and adopting a dual virtual machine support strategy.

Although parallel EVM provides an effective solution, it also brings new security challenges. Parallel execution introduces multi-threaded programming, which increases the complexity of the system. Multi-threaded programming is prone to race conditions, deadlock, live lock, and starvation, which affect the stability and security of the system. At the same time, it introduces new security risks, such as malicious transactions that may exploit the system's parallel execution mechanism to create data inconsistencies or launch competitive attacks.

The article comes from the Internet:Summarizing nine parallel EVM studies: What are the focus of discussion?

Related recommendations: Initia Research Report: Reconstructing multi-chain experience, ecological projects are becoming increasingly mature

Initia redefines the multi-chain experience by seamlessly integrating Layer 1 and Layer 2 solutions into a user-friendly ecosystem. Author: Silent Validator ⚛ Translator: Xiaobai Navigation Coderworld Introduction Initia is a multi-chain Rollups network designed to…