Huobi Growth Academy|Web3 Parallel Computing In-depth Research Report: The Ultimate Path to Native Scaling

1. Introduction: Scaling is an eternal proposition, and parallelism is the ultimate battlefield

Since the birth of Bitcoin, the blockchain system has always faced an unavoidable core problem: scaling. Bitcoin processes less than 10 transactions per second, and Ethereum struggles to break through the performance bottleneck of tens of TPS (transactions per second), which is particularly cumbersome in the traditional Web2 world, which is often tens of thousands of TPS. More importantly, this is not a simple problem that can be solved by "adding servers", but a systemic limitation deeply embedded in the underlying consensus and structural design of the blockchain - that is, the impossible triangle of the blockchain where "decentralization, security, and scalability" cannot be combined.

Over the past decade, we've seen countless expansion attempts rise and fall. From the Bitcoin scaling war to the Ethereum sharding vision, from state channels and plasma to rollups and modular blockchains, from off-chain execution in Layer 2 to structural refactoring of Data Availability, the entire industry has embarked on a path of scaling full of engineering imagination. As the most widely accepted scaling paradigm, rollup has achieved the goal of significantly increasing TPS while reducing the execution burden of the main chain and preserving the security of Ethereum. But it doesn't touch the real limits of the blockchain's underlying "single-chain performance", especially at the execution level, which is the throughput of the block itself – is still limited by the ancient processing paradigm of on-chain serial computation.

Because of this, in-chain parallel computing has gradually entered the industry's field of vision. Different from off-chain scaling and cross-chain distribution, intra-chain parallelism attempts to completely reconstruct the execution engine while maintaining the atomicity and integrated structure of a single chain, and upgrades the blockchain from a single-threaded mode of "serial execution of one transaction by one" to a high-concurrency computing system of "multi-threading + pipeline + dependency scheduling" under the guidance of modern operating system and CPU design. Such a path may not only achieve a hundredfold increase in throughput, but also may become a key prerequisite for the explosion of smart contract applications.

In fact, in the Web2 computing paradigm, single-threaded computing has long been eliminated by modern hardware architectures, and replaced by an endless stream of optimization models such as parallel programming, asynchronous scheduling, thread pools, and microservices. Blockchain, as a more primitive and conservative computing system with extremely high requirements for certainty and verifiability, has never been able to make full use of these parallel computing ideas. This is both a limitation and an opportunity. New chains such as Solana, Sui, and Aptos are the first to start this exploration by introducing parallelism at the architectural level. Emerging projects such as Monad and MegaETH have further elevated on-chain parallelism to breakthroughs in deep mechanisms such as pipeline execution, optimistic concurrency, and asynchronous message-driven, showing characteristics that are getting closer and closer to modern operating systems.

It can be said that parallel computing is not only a "performance optimization method", but also a turning point in the paradigm of blockchain execution model. It challenges the fundamental patterns of smart contract execution and redefines the basic logic of transaction packaging, state access, call relationships, and storage layout. If rollup is "moving transactions to off-chain execution", then on-chain parallelism is "building supercomputing cores on-chain", and its goal is not to simply improve throughput, but to provide truly sustainable infrastructure support for future Web3 native applications (high-frequency trading, game engines, AI model execution, on-chain social, etc.).

After the rollup track gradually tends to be homogeneous, intra-chain parallelism is quietly becoming the decisive variable of the new cycle of Layer 1 competition. Performance is no longer just "faster", but the possibility of being able to support an entire heterogeneous application world. This is not only a technical race, but also a paradigm battle. The next generation of sovereign execution platforms in the Web3 world is likely to emerge from this intra-chain parallel wrestling.

2. Panorama of the expansion paradigm:

As one of the most important, sustainable and difficult topics in the evolution of public chain technology, five types of routes, each with its own focus on expansion, has given birth to the emergence and evolution of almost all mainstream technology paths in the past decade. Starting from the battle over the block size of Bitcoin, this technical competition on "how to make the chain run faster" finally divided into five basic routes, each of which cuts into the bottleneck from a different angle, with its own technical philosophy, landing difficulty, risk model and applicable scenarios.

the first route is the most straightforward on-chain scaling, which represents ways to increase the block size, shorten the block time, or improve the processing power by optimizing the data structure and consensus mechanism. This approach has been the focus of the Bitcoin scaling debate, giving rise to "big block" forks such as BCH and BSV, and also influencing the design ideas of early high-performance public chains such as EOS and NEO. The advantage of this kind of route is that it retains the simplicity of single-chain consistency, which is easy to understand and deploy, but it is also very easy to touch the systemic upper limit such as centralization risk, rising node operating costs, and increased synchronization difficulty, so it is no longer the mainstream core solution in today's design, but has become more of an auxiliary collocation of other mechanisms.

The second type of route is off-chain scaling, which is represented by state channels and sidechains. The basic idea of this type of path is to move most of the transaction activity off-chain, and only write the final result to the main chain, which acts as the final settlement layer. In terms of technical philosophy, it is close to the asynchronous architecture of Web2 - try to leave heavy transaction processing at the periphery, and the main chain does minimal trusted verification. Although this idea can theoretically be infinitely scalable, the trust model, fund security, and interaction complexity of off-chain transactions limit its application. For example, although Lightning Network has a clear positioning of financial scenarios, the scale of the ecosystem has never exploded. However, multiple sidechain-based designs, such as Polygon POS, not only have high throughput, but also expose the disadvantages of difficult inheritance of the security of the main chain.

The third type of route is the most popular and widely deployed Layer 2 rollup route. This method does not directly change the main chain itself, but scales through the mechanism of off-chain execution and on-chain verification. Optimistic Rollup and ZK Rollup have their own advantages: the former is fast to implement and highly compatible, but it has the problems of challenge period delay and fraud proof mechanism; The latter has strong security and good data compression capabilities, but it is complex to develop and lacks EVM compatibility. No matter what type of rollup it is, its essence is to outsource execution power, while keeping data and verification on the main chain, achieving a relative balance between decentralization and high performance. The rapid growth of projects such as Arbitrum, Optimism, zkSync, and StarkNet proves the feasibility of this path, but it also exposes medium-term bottlenecks such as excessive reliance on data availability (DA), high costs, and fragmented development experience.

The fourth type of route is the modular blockchain architecture that has emerged in recent years, such as Celestia, Avail, EigenLayer, etc. The modular paradigm advocates the complete decoupling of the core functions of the blockchain - execution, consensus, data availability, and settlement - by multiple specialized chains to complete different functions, and then combine them into a scalable network with a cross-chain protocol. This direction is strongly influenced by the modular architecture of the operating system and the concept of cloud computing composability, which has the advantage of being able to flexibly replace system components and greatly improve efficiency in specific areas such as DA. However, the challenges are also very obvious: the cost of synchronization, verification, and mutual trust between systems after module decoupling is extremely high, the developer ecosystem is extremely fragmented, and the requirements for medium- and long-term protocol standards and cross-chain security are much higher than those of traditional chain design. In essence, this model no longer builds a "chain", but builds a "chain network", which puts forward an unprecedented threshold for the overall architecture understanding and operation and maintenance.

The last type of route, which is the focus of the subsequent analysis in this paper, is the intra-chain parallel computing optimization path. Unlike the first four types of "horizontal splitting", which mainly carry out "horizontal splitting" from the structural level, parallel computing emphasizes "vertical upgrading", that is, the concurrent processing of atomic transactions is realized by changing the architecture of the execution engine within a single chain. This requires rewriting the VM scheduling logic and introducing a complete set of modern computer system scheduling mechanisms, such as transaction dependency analysis, state conflict prediction, parallelism control, and asynchronous calling. Solana is the first project to implement the concept of parallel VM into a chain-level system, which realizes multi-core parallel execution through transaction conflict judgment based on the account model. The new generation of projects, such as Monad, Sei, Fuel, MegaETH, etc., further try to introduce cutting-edge ideas such as pipeline execution, optimistic concurrency, storage partitioning, and parallel decoupling to build high-performance execution cores similar to modern CPUs. The core advantage of this direction is that it does not need to rely on the multi-chain architecture to achieve a breakthrough in the throughput limit, and at the same time provides sufficient computing flexibility for the execution of complex smart contracts, which is an important technical prerequisite for future application scenarios such as AI Agent, large-scale chain games, and high-frequency derivatives.

Looking at the above five types of scaling paths, the division behind them is actually the systematic trade-off between performance, composability, security, and development complexity of blockchain. Rollup is strong in consensus outsourcing and secure inheritance, modularity highlights structural flexibility and component reuse, off-chain scaling attempts to break through the bottleneck of the main chain but the trust cost is high, and intra-chain parallelism focuses on the fundamental upgrade of the execution layer, trying to approach the performance limit of modern distributed systems without destroying the consistency of the chain. It is impossible for each path to solve all problems, but it is these directions that together form a panorama of the Web3 computing paradigm upgrade, and also provide developers, architects, and investors with extremely rich strategic options.

Just as the operating system has shifted from single-core to multi-core and databases have evolved from sequential indexes to concurrent transactions, the expansion of Web3 will eventually move towards a highly parallel execution era. In this era, performance is no longer just a chain speed race, but a comprehensive embodiment of the underlying design philosophy, depth of architecture understanding, software and hardware collaboration, and system control. And intra-chain parallelism may be the ultimate battlefield of this long-term war.

3. Parallel computing classification graph: five paths from account to instructionIn

the context of the continuous evolution of blockchain scaling technology, parallel computing has gradually become the core path for performance breakthroughs. Different from the horizontal decoupling of the structure layer, network layer or data availability layer, parallel computing is a deep mining at the execution layer, which is related to the lowest logic of the operation efficiency of the blockchain, and determines the response speed and processing capacity of a blockchain system in the face of high concurrency and multi-type complex transactions. Starting from the execution model and reviewing the development of this technology lineage, we can sort out a clear classification map of parallel computing, which can be roughly divided into five technical paths: account-level parallelism, object-level parallelism, transaction-level parallelism, virtual machine-level parallelism, and instruction-level parallelism. These five types of paths, from coarse-grained to fine-grained, are not only the continuous refinement process of parallel logic, but also the path of increasing system complexity and scheduling difficulty.