Author: ChainFeeds
EVM++ utilizes AI models to accurately predict transaction dependencies, reducing parallel execution conflicts, and maintains dApp performance through elastic block space.
In March of this year, the scalability L1 blockchain network Artela introduced EVM++, an upgrade to the next generation EVM execution layer technology. The first “+” in EVM++ represents “Extensibility”, achieved through Aspect technology to support developers in creating custom on-chain programs in a WebAssembly (WASM) environment. These programs can collaborate with EVM to provide high-performance customized application-specific extensions for dApps. The second “+” represents “Scalability”, enhancing network processing capacity and efficiency significantly through parallel execution technology and elastic block space design.
Yesterday, Artela released a whitepaper detailing how it enhances blockchain scalability by developing parallel execution stacks and introducing elastic block space based on elastic computing.
The Importance of Parallel Processing
In the traditional Ethereum Virtual Machine (EVM), all smart contract operations and state transitions must remain consistent across the entire network. This requires all nodes to execute the same transactions in the same order, even if some transactions have no actual dependencies between them. This sequential processing not only leads to unnecessary waiting but also inefficiencies.
Parallel processing allows multiple processors or computing cores to execute multiple tasks or process data simultaneously, significantly improving processing efficiency and reducing runtime, especially for complex or large-scale computing problems that can be decomposed into independent tasks. Parallel EVM is an extension or improvement upon the traditional Ethereum Virtual Machine, enabling the simultaneous execution of multiple smart contracts or contract function calls, significantly increasing the network’s throughput and efficiency. Additionally, it can optimize efficiency in single-threaded execution. The most direct advantage of parallel EVM is enabling existing decentralized applications to achieve internet-level performance.
Artela Network and EVM++
Artela is an L1 blockchain that enhances EVM’s extensibility and performance through the introduction of EVM++. EVM++ is an upgrade to EVM’s execution layer technology, combining EVM’s flexibility with WASM’s high-performance features. This enhanced virtual machine supports parallel processing and efficient storage, enabling more complex and high-performance applications to run on Artela. EVM++ supports not only traditional smart contracts but also dynamically adding and running high-performance modules on-chain, such as AI agents. These agents can run independently as on-chain coprocessors or directly participate in on-chain games, creating truly programmable NPCs.
Artela ensures that network node computing capacity can flexibly expand according to demand through parallel execution design. Furthermore, validator nodes support horizontal scaling, automatically adjusting the scale of computing nodes based on current load or demand. This scaling process is coordinated by elastic protocols to ensure sufficient computing resources in the consensus network. Through elastic computing, Artela guarantees scalable network node computing power, ultimately achieving elastic block space, allowing large dApps to request independent block space based on specific needs. This not only meets the need for expanding public block space but also ensures the performance and stability of large applications.
Artela’s Parallel Execution Architecture Explained
Predictive Optimistic Execution
Predictive optimistic execution is one of Artela’s core technologies and a distinguishing feature from other parallel EVMs like Sei and Monad. Optimistic execution refers to a parallel execution strategy assuming no conflicts between transactions at the initial state. In this mechanism, each transaction maintains a private state version, recording modifications but not immediately finalizing them. After transaction execution, a verification phase checks for conflicts caused by global state changes from other parallel transactions during the same period. Upon detecting a conflict, the transaction is re-executed. Predictive refers to using specific AI models to analyze historical transaction data to predict dependencies between transactions about to be executed, grouping transactions based on potential data access overlap and scheduling their execution order to reduce conflicts and duplicate executions. In contrast, Sei relies on developers defining transaction dependencies files in advance, while Monad uses compiler-level static analysis to generate transaction dependencies files, both lacking EVM equivalence and Artela’s adaptive capability of dynamic predictive models based on AI.
Async Preloading Technology
Async preloading technology aims to address input/output (I/O) bottlenecks caused by state access to improve data access speed and reduce transaction execution wait times. Before transaction execution, Artela preloads necessary state data from slow storage (such as hard drives) to fast storage (like memory) based on predictive models. By preloading required data, it reduces I/O wait times during execution. When data is preloaded and cached, multiple processors or execution threads can access this data simultaneously, further increasing parallelism in execution.
Parallel Storage
With the introduction of parallel execution technology, transaction processing can be parallelized. However, if the speed of data read/write and updates does not synchronize, it becomes a critical factor limiting overall system performance, shifting bottlenecks to the storage layer. Solutions like MonadDB and SeiDB have begun focusing on optimizing the storage layer. Artela has developed parallel storage by integrating various mature traditional data processing technologies to further enhance efficiency in parallel processing.
The parallel storage system is designed to address two main issues: parallelizing storage processing and improving the efficient recording of data states in the database. Common problems in data storage include data inflation during writing and increased pressure on database processing. To effectively address these issues, Artela adopts a separation strategy of State Commitment (SC) and State Storage (SS). This strategy divides storage tasks into two parts: one for quick operations without retaining complex data structures to save space and reduce data duplication, and the other for recording all detailed data information. Additionally, to avoid performance impact when processing large amounts of data, Artela employs a method of merging small data blocks into larger blocks, reducing complexity in data storage.
Elastic Block Space (EBS)
Artela’s Elastic Block Space (EBS) is designed based on elastic computing concepts, automatically adjusting the number of transactions accommodated in blocks according to network congestion levels.
EBS dynamically adjusts block resources based on specific dApp requirements, providing independent scaling block space for high-demand dApps to address significant performance differences in applications’ blockchain performance needs. EBS’s core advantage lies in “predictable performance”, providing predictable transactions per second (TPS) for dApps. Therefore, regardless of whether the public block space is congested, dApps with independent block space will receive stable TPS. Furthermore, if the contracts written by dApps support parallelism, they can achieve even higher TPS. EBS offers a more stable environment compared to traditional blockchain platforms like Ethereum and Solana, which often experience decreased dApp performance during network congestion, such as during token frenzies or DeFi spikes. Artela effectively resolves such issues through customized and optimized resource management.
In conclusion, Artela achieves highly scalable and predictable network performance through parallel execution stacks and elastic block space. This parallel execution architecture accurately predicts transaction dependencies using AI models, reducing conflicts and duplicate executions. Additionally, large applications can access dedicated processing power and resources as needed, ensuring stable performance even under high network loads. This enables the Artela network to support more complex scenarios, such as real-time big data processing and complex financial transactions.