技术解析Artela：为什么“并行EVM”关乎以太坊EVM生态的存续之战？

最近，Paradigm领投了Monad一轮2.25亿美元的巨额融资，引发了市场对“并行EVM”的关注。这引发了对“并行EVM”解决了什么问题的讨论，以及发展并行EVM的瓶颈和关键。在我看来，“并行EVM”是以太坊EVM生态存续之战的关键一环。为什么呢？让我来解释一下： 由于以太坊EVM虚拟机只能串行处理交易，这导致了EVM-Compatible layer1链以及EVM兼容的layer2链都受制于性能，因为它们都基于同一套框架进行状态和交易处理。 相比之下，像Solana、Sui、Aptos等主打高性能的layer1链天生具有并行处理的优势。因此，想要对抗这些高性能layer1链的挑战，EVM基因的链就必须具备并行处理的能力。那么如何做到呢？让我们以新兴的并行EVM链Artela Network为例来详细说明： 1）Artela等强化型EVM layer1链在高度兼容EVM的基础上提高了TPS，并赋予了交易并行处理能力。这些独立的并行EVM layer1链具有独立的共识机制和技术特性，但仍以兼容并扩展EVM生态为目标，相当于通过“换血统”的方式重构EVM链，以服务于EVM生态； 2）类似Eclipse、MegaETH等的扩容型layer2 EVM兼容链，利用layer2链独立的共识和交易预处理能力，在将大量交易批处理到主网之前对交易状态进行筛选和处理，同时还可以选择其他执行层来最终确定交易状态。这种方法将EVM抽象成一个可插拔的执行模块，实现了并行处理能力；但这类方案服务于EVM，却超出了EVM框架的范畴； 3）像Polygon、BSC等的等效Alt-layer1链在一定程度上实现了EVM的并行处理能力，但只是通过算法层的优化，而并没有进行深层次的共识和存储层优化。因此，这种并行能力更多地是一个特定的功能，而未能彻底解决EVM的并行问题； 4）Aptos、Sui、Fuel等的差异型Non-EVM并行链在某种程度上并非EVM链，而是通过某种方式实现了与EVM环境的兼容。这类Non-EVM链的并行能力与EVM链的接轨存在一定问题。 这些方案各有侧重点，比如有并行处理能力的layer2侧重于模块化组合执行层链的灵活性；而EVM-Compatible链着重于特定功能的定制特性；再者，非EVM链的EVM兼容特性更多地追求以太坊的流动性。对于彻底巩固EVM生态并从底层改变并行处理能力的目标而言，强化型EVM layer1赛道显得更为关键。 那么，构建强化型并行EVM layer1链的关键在于什么？如何实现EVM链的重构同时又服务于EVM生态？关键点在于两点： 1）改进state I/O磁盘读取和输出信息的能力。除了简单的交易排序和调度，为了真正提高并行处理能力，需要引入缓存、数据切片甚至分布式存储技术，从根本上平衡读取速度和可能的状态冲突； 2）拥有高效的网络通信、数据同步、算法优化、虚拟机强化，以及将计算和IO任务分离等共识机制层的各种组件优化，从底层组件架构、协作流程等各个方面综合优化和提升，最终实现快速响应、可控计算消耗、高准确性的交易并行处理能力。 具体到并行EVM layer1链项目本身，要采取哪些技术创新和框架优化来实现“并行EVM”呢？ 为了从底层架构层究竟实现资源协调和优化的并行EVM能力，Artela引入了弹性计算和弹性区块空间。弹性计算使网络能够根据需求和负载动态分配和调整计算资源；弹性区块空间能够基于网络中的交易数量和数据大小进行动态调整区块大小。整个弹性设计的工作原理类似于商场自动感应人流量来自动调整扶梯运行，这种模式非常合理。 如前所述，State I/O磁盘读取性能对并行处理至关重要。Polygon、BSC等EVM-Compatible链通过算法实现的“并行”能力可以提高效率2-4倍，但这只是算法层面的优化，其共识层和存储层并没有进行深层次的优化。这引发了以下问题：真正的深层优化是怎样的？ 对此，Artela借鉴了数据库技术方案，在状态读取和写入方面进行了提升。在写入状态方面，它引入了写入前日志（WAL）技术，即当状态发生改变时，先将更改记录写入日志并提交到内存，完成“写入”操作。这实现了操作的异步化，避免了立即进行磁盘写入操作，从而降低了对磁盘的I/O操作。在状态读取方面，采用了相同的异步操作，通过预加载策略提高了读取效率，根据合约历史执行记录来预测下一次特定合约调用可能要读取的状态，并将其预先加载到内存中，从而提高了磁盘I/O请求的效率。 总的来说，这种算法通过内存空间换取执行时间，从根本上提升了EVM虚拟机的并行处理能力，从根本上优化了状态冲突问题。 此外，Artela引入Aspect模块化编程能力，以更好地管理复杂性并提高开发效率。通过引入WASM编码解析以增强编程灵活性；同时，具备底层API访问权限，实现了执行层的安全隔离。这使得开发者可以在Artela的环境下高效地开发、调试和部署智能合约，激发开发者群体的定制扩展能力。特别是，开发者被激励在智能合约代码层面朝着可并行处理的方向进行代码优化，因为减少状态冲突概率对于每个智能合约的调用逻辑和算法都至关重要。 总的来说，“并行EVM”概念实质上是优化交易状态执行过程的过程。Monad号称每秒可处理10,000笔交易，其技术内核主要集中在专用数据库、开发友好度、延迟执行共识、超标量流水线技术等方面，以实现大规模交易的并行处理，这与Artela的弹性计算和I/O异步操作的本质逻辑并无太大差异。 但实际上，这类高性能并行EVM链是在融合web2产品和技术力之后的产物，确实吸收了web2应用市场中高负载状态下的“技术处理”精髓。 在遥远的Mass Adoption未来，实现基于EVM生态的“并行EVM”将成为面向更广泛web2市场的基础基础设施，资本市场对此的看好也是情理之中。

Recently, a huge round of financing of $100 million has attracted the market's attention to parallelism, which has led to a discussion on what problems parallelism has solved and the bottleneck and key to developing parallelism. In my opinion, parallelism is a key link in the battle for the ecological survival of Ethereum. Why? Let me explain that because Ethereum virtual machines can only process transactions in series, the chains and compatible chains are subject to performance, because they are all based on the same framework for state and transaction processing, compared with the priority. The energy chain has the advantage of parallel processing, so the gene chain must have the ability of parallel processing in order to meet the challenge of these high-performance chains. So how to do it? Let's take the emerging parallel chains as an example to explain in detail that the equal-enhanced chains have improved and given the ability of parallel processing of transactions on the basis of high compatibility. These independent parallel chains have independent consensus mechanisms and technical characteristics, but they still aim at compatibility and expansion of ecology, which is equivalent to restructuring the chain by changing bloodlines to serve. The expandable compatible chain with ecological similarity uses the consensus of chain independence and transaction preprocessing ability to screen and process the transaction status before batch processing a large number of transactions to the main network, and at the same time, other execution layers can be selected to finally determine the transaction status. This method abstracts it into a pluggable execution module and realizes the parallel processing ability, but this kind of scheme serves beyond the scope of the framework, and achieves the parallel processing ability to a certain extent, but only through the optimization of the algorithm layer. However, there is no deep consensus and storage layer optimization, so this kind of parallel ability is more a specific function and the parallel problem that has not been completely solved. The differential parallel chain is not a chain to some extent, but is compatible with the environment in some way. There are some problems in the connection between the parallel ability of this kind of chain and the chain. These schemes have their own emphasis, such as the flexibility of modular combination execution layer chain and the customization of specific functions. Moreover, the compatibility of non-chain and the pursuit of the liquidity of Ethereum are more critical to the goal of completely consolidating the ecology and changing the parallel processing ability from the bottom. Then the key to building an enhanced parallel chain lies in what to achieve the reconstruction of the chain and serve the ecology at the same time. The key point lies in two aspects: improving the ability of reading and outputting information on the disk, in addition to simple transaction sequencing and scheduling, in order to truly improve the parallel processing ability, cache data slices and even distribution need to be introduced. Storage technology fundamentally balances the reading speed and possible state conflicts, has efficient network communication data synchronization algorithms, optimizes virtual machines, and optimizes various components of the consensus mechanism layer, such as calculation and task separation, from various aspects such as the underlying component architecture and collaborative process, and finally realizes rapid response. Controllable calculation consumes high-accuracy transaction parallel processing capacity. Specifically, what technical innovations and framework optimization should be adopted for the parallel chain project itself to achieve parallelism? The parallel ability of resource coordination and optimization in the layer architecture layer introduces elastic calculation and elastic calculation of elastic block space, which enables the network to dynamically allocate and adjust the calculation resources according to the demand and load. The elastic block space can dynamically adjust the block size based on the number of transactions and the size of data in the network. The working principle of the whole elastic design is similar to that of a shopping mall automatically sensing the flow of people to automatically adjust the escalator operation. This mode is very reasonable. As mentioned above, the disk reading performance is crucial to parallel processing. The parallel ability of the important equal chain realized by the algorithm can improve the efficiency by times, but this is only the optimization of the algorithm level. The consensus layer and storage layer have not been deeply optimized, which raises the following questions: What is the real deep optimization? For this, the database technology scheme has been used for reference to improve the reading and writing of the state. In the writing state, it has introduced the log before writing technology, that is, when the state changes, the change record is written into the log and submitted to the memory to complete the writing operation. This has been realized. The asynchronization of operation avoids immediate disk writing operation, thus reducing the operation on disk. The same asynchronous operation is adopted in the aspect of state reading, which improves the reading efficiency through preloading strategy. According to the contract history execution record, the state that may be read in the next specific contract call is predicted and preloaded into memory, thus improving the efficiency of disk request. Generally speaking, this algorithm fundamentally improves the parallel processing ability of virtual machines by exchanging memory space for execution time. The problem of state conflict is fundamentally optimized. In addition, modular programming ability is introduced to better manage complexity and improve development efficiency. By introducing code parsing to enhance programming flexibility and having access rights at the bottom, the security isolation of the executive layer is realized, which enables developers to efficiently develop, debug and deploy smart contracts in a new environment, and stimulates the customization and expansion ability of developers, especially the developers are encouraged to optimize the code in the direction of parallel processing at the smart contract code level. Because reducing the probability of state conflict is very important for the calling logic and algorithm of every intelligent contract, generally speaking, the concept of parallelism is essentially the process of optimizing the execution process of transaction state, which is called that transactions can be processed every second. Its technical core mainly focuses on the development of special database, the delay in executing consensus superscalar pipeline technology and other aspects to realize the parallel processing of large-scale transactions, which is not much different from the essential logic of elastic computing and asynchronous operation, but in fact this kind of high-performance parallelism. Chain is the product of the integration of product and technology, which really absorbs the essence of technology processing under high load in the application market. In the distant future, it will become an infrastructure capital market facing a wider market, and it is reasonable to be optimistic about this. 比特币今日价格行情网_okx交易所app_永续合约_比特币怎么买卖交易_虚拟币交易所平台

注册有任何问题请添加 微信：MVIP619 拉你进入群

文章转载注明 网址：https://netpsp.com/?id=53251