Original title: Death, Taxes, and EVM Parallelization
Original author: Reforge Research
Original source: Reforge Research
Compiled by: Mars Finance, MK
introduce
In today’s computer systems, speedups and efficiency gains are often achieved by completing tasks in parallel rather than sequentially. This phenomenon, aptly called parallelization, is a phenomenon that has emerged with the advent of multi-core processor architectures in modern computers. Traditionally, tasks that were executed step by step are now optimized through a concurrent perspective, which maximizes the performance of the processor. Similarly, in blockchain networks, the principle of multitasking is also applied at the transaction level, although it does not rely on multiple processors, but rather on the collective verification power of many validators on the network. Examples of early implementations include:
In 2015, Nano (XNO) introduced a block grid structure where each account has an independent blockchain, enabling parallel processing of transactions and eliminating the need for network-wide transaction confirmations.
In 2018, a paper on the parallel execution engine of the Block-STM (Software Transactional Memory) blockchain network was published, Polkadot achieved parallelization through a multi-chain architecture, and EOS launched a multi-threaded processing engine.
In 2020, Avalanche introduced a parallel processing consensus mechanism (not EVM c-chain, which is serial), and Solana launched a similar innovative technology called Sealevel.
For EVM, since its birth, its transactions and smart contract execution have been sequential processing. This single-threaded execution design limits the overall throughput and scalability of the system, especially during periods of peak network demand. As network validators face an increasing workload, network speeds inevitably slow down, resulting in higher costs for users who have to bid to prioritize their transactions in a crowded network environment.
The Ethereum community has long explored parallel processing as a solution, beginning with Vitalik’s EIP proposal in 2017. The original goal was to achieve parallelization through shard chains, or shards. However, the rapid development and adoption of L2 rollups, due to their simplicity and the direct scalability benefits they provide, has shifted Ethereum’s focus to a technique now known as danksharding. In danksharding, shards serve primarily as a data availability layer rather than parallel transaction execution. However, given that the full implementation of danksharding is still in progress, attention has turned to several key parallelized L1 networks that are compatible with the EVM, especially Monad, Neon EVM, and Sei.
Given the evolution of software systems engineering traditions and the scalability achievements of other networks, parallel execution of the EVM seems inevitable. While we have strong faith in this transition, the future beyond it is still full of unknowns and hopes. This has profound implications for the world's largest smart contract developer ecosystem, currently valued at over $80 billion. What happens when optimized state access drives gas prices down dramatically to just fractions of a cent? How large will the design space be for application layer developers?
Parallelization is a means, not an end.
Scaling blockchain is a complex multi-dimensional problem, and parallel execution paves the way for the development of key infrastructure, such as blockchain state storage. The main challenge facing parallel EVM projects is not only to achieve simultaneous computation; it is also necessary to optimize state access and modification in a parallel environment. The core issues include:
Ethereum clients and Ethereum itself use different storage data structures (B-tree/LSM-tree vs. Merkle Patricia Trie), and when one data structure is embedded in another, it may cause performance degradation.
During parallel execution, asynchronous input/output (async I/O) capabilities for transactional reads and writes are extremely important; otherwise, processes may become stuck waiting for each other, wasting potential speed improvements.
The increase in computational tasks, such as a large number of SHA-3 hashes or calculation operations, is almost negligible compared to the cost of accessing stored values. To reduce transaction processing time and gas costs, the database infrastructure needs to be improved. This is not just about adopting traditional database architecture as an alternative to raw key-value storage (such as SQL DB). The relational model of EVM state adds unnecessary complexity and overhead, and the cost of 'sload' and 'sstore' operations is also higher than that of basic key-value storage. EVM state only needs to be read and written in points, and write operations occur independently of the end of each block. Therefore, the focus of improvement should be on key areas such as scalability, low-latency reads and writes, efficient concurrency control, state pruning and archiving, and seamless integration with the EVM. For example, Monad is building a custom state database called MonadDB from scratch, which uses the latest kernels to support asynchronous operations and natively implements the patricia trie data structure, both on disk and in memory.
We expect to see further refactoring of the underlying key-value database, as well as significant improvements to a host of supporting infrastructure for blockchain storage capabilities.
Let us once again celebrate the exceptional value of programmable central limit order books (pCLOBs).
As DeFi moves towards higher fidelity levels, central limit order books (CLOBs) are increasingly becoming the dominant approach to trading design. Since their debut in 2017, automated market makers (AMMs) have become a core force in the DeFi space, widely admired for their simplicity and unique liquidity-driving capabilities. AMMs have sparked a revolution in the DeFi world by leveraging liquidity pools and pricing algorithms, becoming the preferred alternative to traditional trading systems such as order books. While CLOBs play a cornerstone role in traditional finance, they have had a difficult struggle when introduced to Ethereum due to the limitations of blockchain scalability.
This design requires a large number of transactions, including the submission, execution, cancellation or modification of each order, each step requiring a new on-chain transaction. Given the early stages of Ethereum's scalability work, the cost of this requirement made CLOBs less suitable for the early days of DeFi, leading to the failure of early versions like EtherDelta. However, despite their popularity, AMMs have inherent limitations. As DeFi matures and attracts more sophisticated traders and institutions, these limitations become more prominent.
After recognizing the advantages of CLOBs, people began to increase efforts to integrate CLOB-based exchanges into DeFi on other blockchain networks with higher scalability. Representative projects such as Kujira, Serum (RIP ☠), Demex, dYdX, Dexalot, and more recently Aori and Hyperliquid aim to provide a better on-chain trading experience than their AMM competitors.
However, aside from projects that focus on specific areas, such as dYdX and Hyperliquid’s focus on perpetual contracts, CLOBs in these alternative networks face their own challenges, including:
Liquidity fragmentation problem: Because the DeFi protocol on Ethereum is highly composable and seamlessly integrated, a strong network effect is formed, making it difficult for CLOBs on other chains to attract sufficient liquidity and trading volume, hindering their adoption and promotion.
Memecoins: Bootstrapping liquidity in on-chain CLOBs requires placing limit orders, which is a more challenging chicken-and-egg problem given new and lesser-known assets like memecoins.
CLOB with blob
As for the L2 layer, existing Ethereum L2 solutions have achieved significant improvements in transaction processing capabilities and costs compared to the Ethereum mainnet, especially after the recent Dencun hard fork. By using lightweight binary large objects (blobs) instead of gas-intensive calldata, transaction costs are significantly reduced. As of April 1, Arbitrum and Optimism had fees of $0.028 and $0.064, respectively, with Mantle having the lowest fee of $0.015, according to growthepie.
This decrease is significant compared to Dencun’s high fees before the hard fork, as calldata previously accounted for 70%-90% of costs. Unfortunately, even though fees have dropped significantly, the post/cancel fee of around $0.01 is still considered too high. For example, institutional traders and market makers often have high order-to-trade ratios, executing only a small number of actual trades among a large number of orders. Even under current L2 fee pricing, large order submissions across different books and their subsequent modifications or cancellations can have a significant impact on the profitability and strategic decisions of institutional players, even if the cost per trade is less than $0.01 .
The pCLOB
With the advent of the parallel EVM, DeFi activity is expected to surge, led by CLOBs that can be implemented on-chain. In particular, programmable central limit order books (pCLOBs) are highly composable because DeFi is inherently highly composable and can interact with a variety of protocols (only limited by gas), resulting in a rich trading portfolio. With this feature, pCLOBs can integrate custom logic into the order submission process, which can be triggered before and after the order is submitted. For example, pCLOB smart contracts can:
Verify order parameters (such as price and quantity) based on predefined rules or market conditions;
Implement real-time risk review to ensure that leveraged transactions have sufficient margin or collateral;
Dynamically calculate fees based on various parameters such as order type, volume, market volatility, etc.
Execute orders based on specific conditions; and at a much lower cost than existing trading models.
The concept of “just-in-time (JIT)” liquidity embodies this advantage well. Liquidity will not be stranded on any single exchange, but will be actively mobilized from elsewhere at the moment of order matching, previously realizing gains on the underlying platform. Who can refuse to get every profit on MakerDAO before looking for trading liquidity? Mangrove Exchange’s innovative “Quote as Code” approach demonstrates this potential, where once a quote is matched, the embedded code is executed, with its primary task being to find the liquidity required by the order taker. Despite challenges, particularly in terms of layer-2 (L2) scalability and cost, parallel EVM significantly enhances the matching engine efficiency of pCLOBs. Today, pCLOB can deploy a parallel matching engine to process orders and perform matching calculations in parallel through multiple "channels." Each channel handles a portion of the order book, removing the constraints of price-time priority and only executing when a match is found. This reduces the latency between order submission, execution and modification, allowing the order book to be updated with optimal efficiency.
For long-tail assets with lower liquidity, AMMs may still be widely used; however, for blue-chip assets, pCLOBs will undoubtedly demonstrate its superiority.
In a discussion with Keone Hon, co-founder and CEO of Monad, he confidently predicted that multiple pCLOBs will attract attention in different high-throughput ecosystems and have a positive impact on DeFi as a whole due to their ability to significantly reduce fees. profound impacts on ecosystems.
Even with these advances alone, we expect pCLOBs to have a huge impact in capital efficiency and lead a new trend in the DeFi world.
We recognize that while more applications are needed, the first priority is...
Existing and new applications must be designed in a way that they can fully exploit the underlying parallelization features.
Most decentralized applications today are not parallelizable, and their interactions with the blockchain are sequential in nature. However, history shows that technologies and applications evolve naturally to take advantage of new technological advances, even if they were not originally designed for them. When the first iPhone was launched, applications designed for it were an example of this. We are in a similar transition period, just like adding multi-core processing power to the blockchain, which will enable even better applications.
The evolution of e-commerce, from displaying online magazine catalogs to forming robust two-sided marketplaces, is a prime example of this transformation. With the implementation of a parallel EVM, we will witness a similar evolution of decentralized applications. This further highlights a key limitation: if applications are not designed for parallelism, then they cannot inherently benefit from the efficiency gains brought by a parallel EVM. Therefore, it is not enough to simply implement parallelism at the underlying infrastructure layer, the application layer also needs to be redesigned to ensure alignment with it.
State contention
Even without changing the applications themselves, we still expect to see 2-4x performance improvements. However, why stop there, especially when the potential for performance improvement is even greater? The key challenge with this shift is that applications need to be fundamentally redesigned to accommodate the subtle differences in parallel processing.
In particular, conflicts arise when multiple transactions in a decentralized application attempt to modify the same state at the same time. Such conflicting transactions need to be processed serially to resolve the problem, but this will offset the advantages of parallelization.
We will not discuss conflict resolution methods in detail here, but the number of potential conflicts faced by application developers depends largely on their own design. Decentralized applications of some popular protocols such as Uniswap were designed and implemented without considering this constraint. 0xTaker, co-founder of Aori, a high-frequency offline order book company for makers, has explored in depth the main state contention problems that will be encountered in the context of parallel processing. For AMM, its peer-to-peer pool model may attract a large number of participants to trade against the same pool at the same time, ranging from a few transactions to more than a hundred transactions may cause state contention, so AMM designers must carefully plan how to allocate and manage liquidity in the pool to maximize its benefits.
Sei's core developer Steven stressed the importance of considering contention in multithreaded development, and noted that Sei is actively exploring the implications of parallelization and its impact on resource utilization.
Predictability of performance
Yilong, co-founder and CEO of MegaETH, also stressed the importance of performance predictability for decentralized applications. Performance predictability means that decentralized applications can execute transactions consistently within a given time, regardless of network congestion or other external factors. One way to achieve this goal is to use specific chains. However, while application-specific chains do provide predictable performance, they sacrifice composability.
Parallelization provides a way to minimize state contention through local fee market experiments.
Aori, co-founder of 0xTaker, said that through advanced parallelism and multi-dimensional fee mechanisms, more deterministic performance can be provided to each application while retaining overall composability.
Solana has an excellent fee market system that is localized, so if multiple users access the same state, they pay a slightly higher fee (influx pricing) rather than competing against each other in a global fee market. This mechanism is particularly beneficial for loosely connected protocols that require performance predictability and composability. Imagine a highway system with multiple lanes and dynamic tolling. During peak hours, it provides dedicated express lanes for vehicles willing to pay higher tolls, ensuring that travel times for these users who prioritize speed and are willing to pay extra are both predictable and fast. At the same time, regular lanes remain open to all vehicles, maintaining the overall connectivity of the entire highway system.
The possibility of imagination
While re-architecting protocols to accommodate underlying parallelism may seem challenging, in DeFi and other areas, the design space has expanded significantly and we can expect to see the creation of new applications that are more complex, efficient, and performant. Focus on use cases that were previously difficult to implement due to performance limitations.
Back in 1995, the only internet plan was $0.10 per 1MB of data downloaded, which encouraged people to be careful about which sites they visited. From that to today’s unlimited use, we’ve seen how people’s behavior patterns have changed and new possibilities have been opened up.
We may return to a user acquisition war similar to the early days of centralized exchanges, where DeFi applications, especially decentralized exchanges, will use referral programs (such as points, airdrops) and better user experience as competitive advantages. We envision a world in which any on-chain game with a reasonable amount of interaction is possible. Hybrid order book-AMMs exist, but it is better to move CLOB sorters directly to the chain as independent nodes and decentralized through governance, which will increase decentralization, reduce latency and enhance composability. Now, on-chain social interaction is completely feasible. Indeed, any scenario involving a large number of people or agents performing a specific activity at the same time is now possible.
In addition to humans, intelligent agents are likely to become increasingly dominant in on-chain transaction flows. AI has been a player in the game for some time, for example in arbitrage and automated trade execution, but their participation is expected to grow exponentially at an unprecedented rate. We believe that all forms of participation on the chain will be enhanced by artificial intelligence to some extent. The latency requirements for agents to conduct transactions will be more critical than we currently imagine.
Ultimately, technological advancement is only a fundamental enabler. Ultimately, the winners will be those who can outperform their peers in onboarding users, building trading volume, and liquidity. The difference is that today, developers have more resources at their disposal.
Cryptocurrency UX sucks… Now, it’s going to get better.
User Experience Unification (UXU) is not only possible, but necessary — and the industry is definitely moving towards achieving it.
Today’s blockchain user experience is fragmented and cumbersome — users need to switch between multiple blockchains, wallets, and protocols, patiently wait for transactions to complete, and risk security vulnerabilities or hacks. The ideal future is one where users can seamlessly interact with their assets securely without having to worry about the underlying blockchain infrastructure. We call the transition from the current fragmented user experience to a unified, seamless experience User Experience Unification (UXU).
At its core, improving blockchain performance, especially through reduced latency and lower fees, can significantly solve user experience issues. Historically, improvements in performance have often had a positive impact on all aspects of our digital user experience. For example, faster internet speeds not only make online interactions seamless, but also drive demand for richer, more immersive digital content. The emergence of broadband and fiber optic technologies has facilitated low-latency streaming of high-definition video and real-time online gaming, raising user expectations for digital platforms. This growing need for depth and quality has spawned continued innovation by companies in developing the next big and sexy thing—from advanced interactive web content to sophisticated cloud-based services to virtual/augmented reality experiences. Increased internet speeds have not only improved the online experience itself, but have also correspondingly expanded the scope of user needs.
Likewise, improved blockchain performance will not only enhance the user experience directly by reducing latency, but will also do so indirectly by enabling the rise of protocols that unify and advance the overall user experience. Performance is a key ingredient to their existence. These networks, especially the parallel EVM, will attract more developers because higher performance and lower gas fees mean that onboarding will be more frictionless for end users. In our conversation with Sergey, co-founder of Axelar, he envisioned a world that is not only truly interoperable, but more symbiotic.
If you have complex logic on a high-throughput chain (e.g., parallel EVM), and the chain itself can “absorb” the complexity and throughput requirements of that logic given its high performance, then you can use an interop solution to export that functionality to other chains in an efficient manner.
As scalability issues are resolved and interoperability between different ecosystems increases, we will witness the emergence of protocols that bring web3 user experience to match web2. Some examples include intent-based protocols v2, advanced RPC infrastructure, chain abstraction enablement, and open computing infrastructure enhanced by artificial intelligence.
The state orchestration of our nodes is accelerated by the increased network throughput, as the solvers can resolve our intent very quickly.
Worth to talk about
As performance requirements increase, the oracle market will become frothy.
Parallel EVM means that performance demands on oracles will increase, an area that has been extremely underdeveloped over the past few years. Increased demand from the application layer will incentivize a complacent market full of inefficient performance and poor security, which is necessary to improve DeFi composability. For example, market depth and volume are two strong indicators for many DeFi primitives such as money markets. We expect large incumbents such as Chainlink and Python to be able to adapt fairly quickly as new players challenge their market share in this new era. After speaking with a senior member of Chainlink, our thoughts are consistent: "If Parallel EVM becomes dominant, we may want to refactor our contracts to benefit from it (e.g. reduce dependencies between contracts so that transactions/calls do not have to unnecessarily depend on each other and therefore cannot be exploited by MEV), but because Parallel EVM is designed to improve transparency and throughput for applications already running on EVM, it should not affect network stability."
Parallel EVM L2 also wants to join in the fun. From a technical perspective, creating a high-performance parallel EVM L2 solution is easier than developing an L1 because the setup of the sorter in L2 is much simpler than the consensus-based mechanisms used in traditional L1 systems such as Tendermint and its variants.
We expect optimistic parallel EVM L2 based rollups to dominate in the near term. Eventually, we do expect a transition from OP-based rollups to zk-rollups via general-purpose zk frameworks like RISC0 rather than the traditional approaches used in other zk-rollups.
Rust's advantages... at least for now. We primarily prefer Reth, the Rust implementation of Ethereum, over any other alternatives. This preference is not random, as Rust has many advantages over other languages, including memory safety without garbage collection, zero-cost abstractions, and a rich type system.
The choice of language will play a major role in the evolution of these systems. We firmly believe that Rust will win out in the end. However, porting one implementation to another is not a simple task. It requires significant resources, time, and expertise, which further emphasizes the importance of choosing the right language from the beginning.
In the context of parallel execution, it is also important to mention Move. Move introduces the concept of "resources", which can only be created, moved, or destroyed, but not copied. This ensures that resources are always uniquely owned, preventing common problems that can arise in parallel execution, such as race conditions and data competition.
Formal Verification and Static Types: Move is a statically typed language with an emphasis on safety. It includes features such as type inference, ownership tracking, and overflow checking that help prevent common programming errors and vulnerabilities. These safety features are especially important in the context of parallel execution, where defects can be more difficult to find and reproduce. The language's semantics and type system are based on linear logic, similar to Rust and Haskell, which makes it easier to reason about the correctness of Move programs, so formal verification can help ensure that concurrent operations are safe and correct.
Move advocates a modular design approach, where smart contracts are composed of smaller, reusable modules. This modular structure can make it easier to reason about the behavior of individual components and can facilitate parallel execution by allowing different modules to execute concurrently.
Future considerations: EVM needs to address its security risks.
While we are optimistic about the prospects of a post-parallel EVM on-chain universe, none of this will matter without addressing the security vulnerabilities of the EVM and smart contracts. Unlike network economics and consensus security, hackers frequently exploit smart contract vulnerabilities in DeFi protocols on Ethereum, with theft exceeding $1.3 billion in 2023. As a result, users tend to use walled CEXs or "decentralized" protocols with centralized validator sets to gain a higher sense of security (and performance) to optimize the on-chain experience.
The lack of inherent security features in the EVM design is the root cause of these security vulnerabilities. Comparing the security measures of blockchain with the strict safety standards of the aviation industry, the latter's high security supports people's attention to the safety of life and property. Thorough testing, redundancy, fault tolerance and strict development standards are the guarantee of the aviation industry's safety record, and these key features are mostly missing in EVM and even other VMs.
One solution is to adopt a dual VM architecture, where a separate VM, such as CosmWasm, monitors the execution of EVM smart contracts in real time, similar to antivirus software in an operating system. This architecture allows for advanced inspections, such as call stack inspections, to reduce hacker attack incidents. However, this requires a major transformation of existing blockchain systems. We expect better solutions such as Arbitrum Stylus and Artela to effectively implement such architectures from the beginning.
Most of the existing security primitives in the market are reactive responses to potential or attempted threats by reviewing mempools or smart contract codes. Although this helps with security, it does not fundamentally address the vulnerabilities of VM design. More proactive efforts and more resources must be invested to comprehensively improve the security of blockchain networks and their application layers.
We advocate for a fundamental overhaul of the blockchain VM architecture to introduce real-time protection and other key security features, modeled after battle-tested and proven industry standards such as aviation, for example, through dual VM architectures. Going forward, we support infrastructure improvements that emphasize preventative measures to ensure that security improvements are matched by industry advances in performance, such as parallel EVMs.
in conclusion
The rise of the Parallel EVM marks a new era in the development of blockchain technology. By enabling parallel execution of transactions and optimized state access, it opens up new horizons for decentralized applications. From the resurgence of programmable CLOBs to the rise of more complex and high-performance applications, the Parallel EVM lays the foundation for a more unified and user-friendly blockchain ecosystem. As the industry gradually accepts this change, we can expect explosive growth in the potential of decentralized technology. Ultimately, the success of this change will depend on the ability of developers, infrastructure providers, and the broader community to adapt to the principles of parallel processing and work together to usher in a future where technology is seamlessly integrated into daily life.
The rise of the parallel EVM has the potential to revolutionize decentralized applications and user experiences. It solves the scalability and performance limitations that have long hindered the development of key verticals such as DeFi, and opens a new path for complex, high-throughput applications to flourish while maintaining a balanced approach to the trilemma. Achieving this vision requires more than just infrastructure advances; developers must fundamentally rethink their application architectures to adapt to the principles of parallel processing, minimize state contention, and maximize performance predictability. Although the future is bright, we emphasize the importance of security while pursuing scalability.