Parallel EVM Track Innovation and Outlook

Author: Artela Chinese Blog

How deep is your understanding of "Parallel EVM"? What changes do various components of the blockchain need to make for transactions to be executed in parallel? If we were to categorize the levels of parallelization, what tiers would there be? Do new directions like PayFi need to be equipped with this new engine?

This article will introduce some in-depth concepts of Parallel EVM, exploring the extreme innovations behind it and discussing the ongoing development and prospects of technology and scenarios.

Non-technical Understanding of Parallel EVM

When the EVM executes a block, the transactions within the block can only line up in a single queue, waiting to be processed one by one. In contrast, Parallel EVM can arrange the transactions in the block into multiple queues, allowing them to wait for execution in parallel, thus enabling faster completion of block execution.

Splitting a single queue into multiple queues, this method seems simple, but achieving it is not an easy task. Even if you run a fast-food restaurant: opening multiple order windows can create multiple queues. However, the trouble that comes with it is: can the kitchen staff keep up with the order volume? Even if you increase the number of chefs, there are still further difficulties due to limited kitchen space and equipment constraints.

Ultimately, you can see that a fast-food restaurant with higher throughput often needs to introduce highly automated equipment to achieve this. The seemingly simple act of opening a few more order windows requires the supply of automated equipment that costs tens of millions of dollars to make it meaningful.

Levels of EVM Parallelization

Similar to running a restaurant, Parallel EVM is easy to understand on the surface, but a deeper look reveals that all components of the blockchain need to cooperate and be transformed to achieve the high-performance effects brought by parallel execution. Making EVM parallelized is not just about solving a single-point problem; it involves addressing a series of interconnected technical issues, requiring optimization from the surface down to the core.

Here we use the concept of "levels of parallelization" to review the optimization depth of different Parallel EVM projects.

Level 1 --- Parallelizing Execution

Grouping the transactions in a block so that they can be executed in parallel by the EVM, while ensuring that the execution results are consistent with those of the original serial execution. This feature is the most basic implementation of Parallel EVM, enhancing the execution speed of blocks.

As a pioneering project exploring Parallel EVM, Polygon PoS and BNB Chain have attempted to introduce parallel execution to enhance performance on operating blockchains. However, since they are all operating blockchains, it is difficult to achieve comprehensive optimization of components, typically only realizing this level of parallelism.

Level 2 --- Optimizing Storage

After the EVM completes the execution of transactions, it needs to update data storage, so the efficiency of data updates also needs to be improved alongside parallel execution; otherwise, storage will become a new bottleneck. This feature elevates Parallel EVM to a new level, improving not only execution speed but also storage efficiency.

New Layer 1 public chains like Sei, Monad, and Artela, unlike Polygon PoS, do not carry historical burdens and are designed around parallelization, thus focusing on optimizing storage levels. For example, Sei DB, Monad DB, and Artela's Parallel Storage are all high-performance storage solutions designed for parallelization.

Level 3 --- Pipelining Consensus

Once execution and storage are no longer bottlenecks, it becomes possible to require blocks to package more transactions faster, which involves improvements to consensus algorithms. This level can further enhance the overall transaction throughput of the blockchain.

One of the novel aspects of Monad is deferred-execution, which separates execution from the consensus process, allowing the consensus process to focus on the data of the block rather than execution, thereby achieving higher throughput consensus. Artela's elastic block space allows for flexible scaling of block size during the consensus process.

Level 4 --- Elastic Computing

When consensus, execution, and storage are all parallelized, the blockchain can enter the next stage: achieving elastic characteristics similar to the cloud.

On the basis of full-stack parallelization, Artela has implemented "elastic block space," which, like the cloud, flexibly expands the actual processing power needed based on the demands of dApp subscriptions. This may represent the ultimate level of parallelization. After dApps enjoy elastic block space, they have predictable performance, ensuring stability in large-scale adoption scenarios aimed at consumers.

We can further understand these four levels of parallelization through Artela's technical diagram, which resembles the evolution path of traditional high-performance servers, evolving from a single server to read-write separation, and then to elastic expansion of both read and write capabilities.

How Will EVM Evolve After Parallelization?

From the levels of parallelization above, it can be seen that Parallel EVM is merely the surface; the deeper the optimization goes, the more it relates to the optimization of other non-EVM components of the blockchain. The core of high-performance and scalable blockchain technology lies behind it. The high performance of EVM public chains revolves around extreme innovations compatible with EVM.

Is there an endpoint to performance optimization of blockchains based on Parallel EVM? We can glean some insights from the development history of server technology in Web2.

With the gradual advancement of large-scale applications, the development of Web2 server technology has gone through several stages:

1. Single Machine Architecture: Early servers relied on a single physical device, lacking flexibility in scaling, with performance limitations.
2. Scale-up: Enhancing the performance of a single server by upgrading hardware (such as stronger CPUs and memory), but single-machine scaling has obvious bottlenecks.
3. Scale-out: In the era of cloud computing, by adding servers and forming clusters, horizontal scaling of computing power is achieved, breaking through the performance limitations of single machines.
4. Heterogeneous Computing: With the surge in AI demand, utilizing heterogeneous computing resources like GPUs and TPUs to enhance processing capabilities for AI and deep learning tasks, meeting the demand for high-performance computing power.

When we place the latest blockchain high-performance computing on the same chart, we can see a technology architecture with future imagination.

New blockchains without historical burdens, such as Sei and Monad, are achieving extreme scale-up through Parallel EVM, while Artela, based on parallelization, is providing predictable performance through scale-out. In addition to Parallel EVM, Dual VMs solutions are also being adopted, with Arbitrum and Artela incorporating WASM as heterogeneous computing into the blockchain architecture to further complement the shortcomings of EVM.

The Significance of Enhancing EVM Parallelization Levels

Returning to the restaurant example, a modern semi-automated high-efficiency kitchen is not suitable for all restaurants, but it can open up the fast-food market. Existing classic DeFi protocols may operate well on any infrastructure; they seem to have little need for the support of a parallelization engine. However, the crypto industry is in a phase of shifting towards value realization, and some new protocols now have a foundation to try.

PayFi is a culmination of crypto assets, stablecoins, payments, and DeFi, representing an innovative application model that combines payment functions with financial services. Through PayFi, blockchain payment systems will truly unleash their potential, promising to create a more liquid and adaptable global financial market. Having fast settlement capabilities, low transaction fees, and a strong developer ecosystem are important conditions for realizing PayFi. Compared to non-EVM high-performance public chains, Parallel EVM has the unique advantage of being compatible with a large existing ecological infrastructure. The EVM ecosystem boasts the strongest TLV, developers, and mature stacks, and Parallel EVM continues to provide fast settlement capabilities and low transaction fees, making it the best incubator for PayFi.

The article by Reforge Research "Death, Taxes, and EVM Parallelization" summarizes the insightful perspectives of major participants in Parallel EVM, revealing possibilities for some new DeFi directions: programmable central limit order books (pCLOB) now have a feasible foundation. By leveraging the high performance and high parallel processing capabilities provided by Parallel EVM, pCLOB can achieve complex and efficient trading functions similar to traditional financial trading platforms in a decentralized environment. It allows developers to embed custom smart contract logic during the submission and execution of orders, which can be used for additional verification, determination of execution conditions, and dynamic adjustment of transaction fees, providing greater flexibility and security, supporting more complex trading strategies and financial products.

The consumer narrative began to emerge during the Token2049 event in Singapore in 2024, where we saw many consumer-oriented projects: for example, Morph launched a consumer-grade Layer 2 with the slogan "A Global Consumer Layer," aiming to bring the value of Web3 into daily life and break the curse of "ghost towns" on public chains and the illusion of false TVL prosperity; or purely to C projects like Matr1x Fire and Shrapnel, which are hardcore FPS games that, in reality, have gameplay experiences not much different from traditional FPS games. These projects have found good angles to push crypto further into the mainstream. Compared to past DeFi and NFT dApps, their common characteristic is the need for large-scale applications, which places higher demands on infrastructure. The Parallel EVM infrastructure can serve as a solid foundation for this.

If all these new directions open up the market, the parallelization capabilities of EVM will become a long-term project of continuous optimization and breakthroughs, akin to cloud computing facilities.

Conclusion

Parallel EVM is merely a summary of high-performance EVM, superficially representing the parallel execution of EVM transactions, but in reality, "pulling one hair affects the whole body" --- it requires deep optimization of technical components behind it. We use levels of parallelization to summarize the different degrees of implementation. The infrastructure is unlocking consumer scenario use cases, and the advantages of Parallel EVM are expected to elevate the past ecological accumulation to a new level.