1. Computation and resource bottlenecks under traditional frameworks
Traditional blockchain technology, represented by Bitcoin and Ethereum, has made remarkable achievements in decentralization, transparency and security, and has promoted the development of encryption technology and applications. However, due to the "blockchain impossible triangle" problem (Figure 1-1), there are obvious bottlenecks in computing performance and resource utilization, which hinders technological innovation and application development, and brings challenges to the encryption industry.
Figure 1-1. Blockchain Impossible Triangle
First, let’s analyze the three elements in the “blockchain impossible triangle”:
Security: Security essentially reflects the consensus requirements, specifically manifested in ensuring the consistency, integrity, tamper-proof nature, traceability, and verifiability of block data. Meeting these characteristics enables blockchain to establish a strong trust security mechanism of 'trustless.' Therefore, the security of consensus is the primary demand of blockchain and the foundation of its development.
Decentralization: Decentralization means that there is no single control point in the system, with power and control distributed across multiple nodes, which can enhance the system's fault tolerance, resistance to censorship, and security, preventing single points of failure and malicious manipulation. While distributed systems may not necessarily be decentralized systems (for example, a distributed system controlled by a single entity is not decentralized), decentralized systems are always distributed.
Scalability: In the concept of the 'blockchain impossibility triangle,' scalability refers to the computational performance scalability of distributed systems. For digital systems, everything is computation, and different applications have varying computational performance requirements. Broadly speaking, scalability refers to the system's ability to handle an ever-growing amount of data, transaction volume, and user count, which is reflected not only in TPS (transactions per second) but also in storage capacity, network bandwidth, and the number of nodes. High scalability is necessary to support large-scale applications and user growth. The scalability of distributed systems directly affects the innovation and scaling of decentralized applications (DApps) upon them.
Among the above three elements, blockchain emphasizes decentralization, strengthens verification and consensus security, while being relatively weak in computational performance. This leads to the blockchain impossibility triangle dilemma: when the needs for decentralization and consensus security are met, the scalability of computation will be restricted, as seen in Bitcoin. This means that within such a system framework, the distributed systems of blockchain struggle to support application innovations requiring high computational performance or cannot meet the scaling demands of applications, such as AI big data models, graphics rendering, on-chain gaming, and large-scale social interactions.
The above primarily analyzes the computational performance expansion challenges posed by the blockchain impossibility triangle. Where does the root of this issue lie? Next, we will explore the interrelationships among the elements within the formation process of a block.
In blockchain technology, a 'block' refers to a dataset formed by packaging a series of verified transaction data within a specific time interval. This concept includes the following key elements and their interrelations:
Consensus (Data): Verified transaction data with consistent state, i.e., consensus data formed within the block.
Block Space: Refers to the storage space for transaction data. These transactions are encapsulated within blocks, and the number of transactions that can be stored is limited by the block size (as determined by the system or restricted by the total gas fees of the block), meaning the on-chain storage space is a limited resource, which in turn affects the scalability of applications.
Computational Performance: The number of transactions packed divided by block time gives the number of transactions processed per second, i.e., TPS (Transactions Per Second) = Number of transactions in the block / Block time. Computational performance is correlated with the consensus process and storage space.
From the above analysis, it can be seen that the three elements of consensus, storage space, and computational performance in a block are interrelated, forming a restrictive relationship. While pursuing consensus consistency, blockchain not only constrains the scalability of individual block storage space but also limits the scalability of computational performance. This is precisely the root cause of the blockchain impossibility triangle problem.
Further analysis indicates that during the block formation process, blockchain systems construct three global, system-level resources: data (consensus) resources, storage resources, and computing resources. However, the impossibility triangle problem limits the effectiveness and scalability of these three resources, creating resource bottlenecks that hinder their full potential. If there were a way to break this constraint, would it bring about a resource-driven new development landscape for blockchain?
This is exactly the core issue that this paper seeks to address. Research shows that from the SCP paradigm, super-parallel computing model Actor to SSI distributed system architecture, a complete technological chain has formed in the engineering practice of AO + Arweave, breaking the blockchain impossibility triangle dilemma, fully releasing the resource potential of blockchain and distributed systems, and providing empowerment in practice, thus opening a new development path for value creation and scalable applications in Web3.
II. SCP: Breaking Through the Bottlenecks of Computational Performance and Resource Expansion
2.1. Breaking the Blockchain Impossibility Triangle Based on SCP
AO (Super Parallel Computing Network) is built on Arweave, realizing the engineering application of the Storage-based Consensus Paradigm (SCP). As shown in the figure below:
Figure 2-1. Modular System Architecture of AO + Arweave Based on SCP
Based on the core concept of SCP, the AO + Arweave system architecture effectively separates on-chain storage (consensus) from off-chain computation:
At the storage level, the storage resources provided by Arweave are responsible for the permanent storage of data, and blockchain technology ensures the traceability and immutability of on-chain data, achieving consistency and high availability of data, reflecting the concept of 'storage is consensus.'
At the computing layer, computing tasks are migrated off-chain and decoupled from the storage (consensus) layer. This design allows computational performance to be free from direct constraints of on-chain consensus, enabling infinite scalability through the addition of off-chain computing nodes, greatly enhancing processing efficiency and system flexibility.
Overall Effect: Arweave's storage public chain maintains the decentralization of the system and the consensus security of data, while AO ensures the infinite scalability of computing performance off-chain. This structure ensures that the needs for decentralization, consensus security, and computational performance scalability are met within the entire AO + Arweave system, effectively addressing the challenges of the blockchain impossibility triangle.
2.2. Constructing Three Types of Global System-Level Resources
The above features based on SCP play important roles in the operational practice of the system; they allow storage, computation, and data (consensus) to be both interconnected and independently existing system elements, forming global, system-level resources, as shown in Figure 2-2:
Figure 2-2. Global system-level resources in the AO network
Storage Space Resources: As a storage public chain, Arweave's storage space expansion is not limited by block size or total gas fees, but is entirely determined by storage demand, achieving true infinite scalability. This not only meets the system's need for flexible storage space but also enriches the diversity of on-chain data types, providing more possibilities for the innovation of on-chain native applications.
Computing Resources: The AO computing network consists of MU, SU, and CU; here we will first discuss CU, and later analyze the roles and interrelations of each network unit in detail. CU is the unit responsible for computation and can horizontally scale to form CU clusters. These clusters compete for computing rights, supporting different processes running in different CUs in parallel. This design of scalability and parallelism enables AO to provide infinite computing node resources, supporting high-performance parallel computing.
Data (Consensus) Resources: On Arweave, any type and size of data can be permanently stored in the form of 'atomic assets', such as NFTs, documents, images, audio and video, webpages, games, legal contracts, program code, etc. This data forms a tamper-proof massive database, laying the groundwork for data monetization and circulation. Additionally, AO does not achieve consensus on the state of the computation itself but focuses on ensuring that interaction logs are written to Arweave, ensuring the persistent availability and integrity of data, and ensuring the consistency and verifiability of computation output results. Any type of data can be referenced without permission or trust, achieving new value creation.
Security Resources: In fact, during the operation of AO, security resources supported by the protocol token $AO are also built, but this is not directly related to SCP; it pertains to the operation and security mechanisms of AO network communication units and will be analyzed in detail in Section 3 'Customizable Security and Security Resources.'
2.3. Trusted Computer Based on Storage Consensus
Utilizing the above system-level resources and distributed characteristics, AO is built on Arweave's storage public chain, forming a cloud computing network. Similar to traditional Web2 cloud computing, AO theoretically possesses infinite scalability in computing and storage resources, capable of supporting vast data resources. However, AO's uniqueness lies in its establishment of a decentralized, globally consistent consensus trusted computing platform based on the storage consensus paradigm.
First, Arweave provides a permissionless, permanent storage service for global users, building a consensus data foundation that does not rely on trust.
Secondly, AO stores the source code of various applications on the Arweave chain; this code can be downloaded and run locally. Its inputs come from trusted data on-chain, ensuring the consistency and predictability of output results under fixed inputs and execution logic.
Finally, any client can perform consistency validation because, under the same input parameters and execution logic, its computation output results must be consistent, thus ensuring trustworthiness.
It can be seen that, with determinism in source programs, inputs, and outputs, AO constructs a trusted computing system based on storage consensus.
The storage consensus paradigm differs from typical node consensus systems; in the storage consensus paradigm, computation, verification, and consensus are all conducted off-chain, with the final consensus data submitted on-chain for storage, forming the system's availability layer, consensus layer, and settlement layer. This means that, with the support of SCP, computational performance is no longer constrained by consensus and can scale infinitely off-chain. This mechanism provides feasibility for AO network to build a high-performance computing architecture that supports high parallelism and distribution.
So, how did AO evolve into a decentralized world computer with distributed deployment and high parallel execution? This mainly results from the Actor model, network communication units, and the distributed architecture realized by SSI.
Three, Super Parallel: Actor Model and Network Communication Units
3.1. Defining the Basic Framework of Parallel Computing with the Actor Model
The name of the AO network comes from 'Actor Oriented,' indicating that it is a super-parallel computing network. This designation stems from the core Actor model it employs, which sets the foundational structure for parallel computation in the system.
In the Actor model, an 'actor' is the basic unit of parallel computation, consisting of three elements: State, Behavior, and Mailbox. These three elements and their interactions constitute the core concept of the Actor model, as illustrated in Figure 3-1:
Figure 3-1. Actor Model Schematic (Image source: Reference Material 5)
This model defines the core components and interaction rules of the system; actors can be viewed as independent, concurrently active entities that can receive messages, process messages, send messages, and dynamically create new actors. This model has the following characteristics:
Asynchronous Communication: Multiple actors send uniformly formatted messages between each other in a peer-to-peer manner; the sending and processing of messages occur asynchronously, making this communication method inherently suitable for interactions between nodes in distributed systems.
Parallel Execution: Each actor operates independently, with no shared state, so there is no need to worry that the state of other actors will affect its own; each actor can independently handle its own tasks, achieving true parallel operation.
Distributed Deployment: Actors can be deployed and scheduled to run on different CPUs, nodes, or even different time slices, without affecting the final results.
Scalability: Due to its distributed nature and loosely coupled design, the Actor model can flexibly scale horizontally by adding nodes and dynamic load balancing.
In summary, the Actor model optimizes parallel and concurrent issues with its elegant handling mechanism, making it particularly suitable for building distributed systems and high-concurrency applications. The AO network adopts the Actor model as the foundational architecture for parallel computing, thereby achieving efficient asynchronous communication, parallel execution, distributed deployment, and excellent scalability.
3.2. Efficient Parallel Computing Implementation of Communication Network Units
The Actor model provides a framework for parallel computation, while the communication network units of AO embody the practical implementation of this model. These network units include Message Unit (MU), Scheduling Unit (SU), and Computing Unit (CU), each of which is an independent 'actor' that collaborates and synchronizes through uniformly formatted messages (ANS-104). Figure 3-2 shows the basic functions and message interaction processes of these network units.
Figure 3-2. Working principle of AO network communication units (Image source: AO White Paper)
In the AO network, starting an application will trigger the initiation of one or more processes, with the system configuring resources such as memory, virtual machines, and communication network units for each process. Interactions between processes are completed entirely through messages. First, messages from users or other processes are sent to MU, which then forwards the messages to SU for sorting. The sorted messages and their results are permanently stored on Arweave and are processed by a CU within a competitive computing rights cluster, meaning processes can run on any computing node, showcasing the typical characteristics of decentralized parallel computing. After the computation is complete, CU returns the results to SU in the form of a signed certificate to ensure the accuracy and verifiability of the computation results, which are ultimately uploaded to Arweave by SU. The complete dataset formed by each process—including initial state, processing steps, and final results—will be permanently stored on Arweave, becoming consensus data available for others to retrieve, verify, and use.
Figure 3-3. Communication process among units in Token transfer (Image source: AO White Paper)
Figure 3-3 shows the specific application scenario of the AO network processing Token transfer requests, clearly depicting the composition and communication processes of each modular network unit, as well as the distributed storage mechanism formed by interactions with Arweave.
The AO system comprehensively utilizes computing resources (distributed CU clusters), storage resources (distributed Arweave nodes), and data resources (long-term available data stored in Arweave), laying the foundation for AO to become a global computing platform. Based on the Actor model, AO's computing network not only features asynchronous communication, parallel execution, and distributed deployment but also possesses outstanding scalability, making it a truly decentralized, distributed, and parallel computing network.
3.3. Customizable Security and Security Resources
In the previous section, we explored the composition and working principle of the AO network communication unit. In this section, we will delve into the security of this network, which is closely linked to the native token of the AO protocol, $AO. This analysis will resonate with the 'security resources' content in section 2.2, focusing on customizable security and security resources in the AO network.
The network communication units composed of MU, SU, and CU are the core components of the AO computing network, establishing the operational mechanism of the decentralized world computer, forming three types of system-level resources: computing, storage, and data. These are the foundations of the technical model and resource model in the AO network. Based on the technical and resource models, the AO system has created a demand-driven customizable security mechanism. This is based on the protocol's native token $AO, with economic game theory providing security guarantees, thus establishing a security market in AO.
For ease of understanding, we will simplify the security mechanisms in AO from the user's perspective into several core elements and their interrelations: customization needs, security/economic resources, security mechanisms, and the security competition market.
Figure 3-4. The relationships among elements in the AO network security mechanism.
Figure 3-4 describes the interrelationships among elements in the AO network security mechanism:
Customization Needs: As a super-parallel computing platform, each node in AO operates independently and in parallel on various processes, handling different types of data. These different data transaction scenarios have varying requirements for the system's latency, cost, and efficiency, necessitating that AO's security model be flexible enough to customize security strategies based on needs. Users can customize the specific security level required for each message, thus promoting the customization and effective allocation of security resources.
Security/Economic Resources: $AO is the native token of the protocol, serving as a circulating public value unit and economic resource that supports all security mechanisms' economic gaming mechanisms within the AO network.
Security Mechanisms: In various processes of AO, including nodes like MU, SU, and CU, staking $AO is required to participate in security mechanisms. Through staking economic value, the system manages funds and enforces penalties according to rules to prevent malicious behavior. For instance, if MU signs an invalid message or CU provides invalid signature proof, the system will reduce their staked assets.
Security Competition Market: Since security is purchased per message, different messages correspond to different staking requirements, creating a dynamic competitive market. The price of security is determined by the market's supply and demand rather than fixed network rules. This market competition mechanism promotes the effective pricing and allocation of security resources, providing tailored security solutions.
In summary, the decentralized peer-to-peer market structure of the AO network essentially allows nodes to independently set fees for their message passing services, adapting to the different security level requirements of various data transactions and reflecting the system's efficiency in responding to specific security needs. This flexibility enables it to dynamically adapt to changes in market demand and supply, fostering competition and enhancing response efficiency, thus achieving efficient market equilibrium.
$AO's liquidity serves as a tool for economic gaming, establishing a comprehensive, real-time token valuation framework while simultaneously establishing security mechanisms. A well-established valuation framework and metrics for the $AO token economic model will undoubtedly further enhance the security of the AO network.
IV. SSI: Distributed System Architecture for a Unified Experience
In previous discussions, we have articulated the basic framework provided by the Actor model for parallel computing in the AO network, as well as how the network communication units consisting of MU, SU, and CU specifically implement this model. These communication units are deployed across different heterogeneous nodes in a distributed network, allowing process execution without being constrained by specific physical locations, and enabling seamless user interaction through the network. All of this collectively forms a unified computing environment, achieving a Single System Image (SSI), which is the foundation for AO network’s support of countless processes. This section will explore the definition of SSI and its specific role in AO.
The Single System Image (SSI) is a core concept in distributed computing, integrating physically separated heterogeneous computing resources into a unified resource pool through virtualization technology. This integration not only raises the level of abstraction of the system but also greatly optimizes the user experience. Under the influence of SSI, even though the system may consist of multiple servers, distributed databases, or several networks, users perceive it as operating a single computer.
Typically, the SSI structure includes user layer, unified interface, resource management layer, computing nodes, and storage layer, as illustrated in Figure 4-1.
Figure 4-1. Single System Image (SSI) Structure Schematic
Users interact with the SSI system at the user layer through client or web front-end. The unified interface is responsible for receiving user requests and dispatching these requests to the resource management layer. The resource management layer schedules distributed computing nodes and storage resources to execute parallel computing tasks or perform data read and write operations.
SSI provides a feasible solution to the current multi-chain coexistence issue in public chains. For example, the Ethereum ecosystem, due to rapid growth, faces congestion, inefficiency, and high cost issues, while Layer2 serves as the primary solution to these scalability problems, introducing new challenges. Each Layer2 chain, while redundantly building infrastructure, also leads to liquidity fragmentation and cross-chain asset risks, increasing the complexity and threshold for users switching between chains, severely impacting user experience and the scaling development of applications.
Public chains like Solana and Polkadot have already recognized these issues and are making adjustments based on their original architectures. However, AO adopted the distributed architecture of SSI from the start, demonstrating foresight and prescience.
Utilizing the Actor model, the network communication units of AO are hosted on a heterogeneous node set in a distributed network, with these nodes potentially distributed across various global locations, including various types and functions of servers. The AO computing network based on the Actor model is a decentralized distributed network that requires a unified architecture for integration to provide consistent availability and user experience.
When users initiate an AO process through the front end, the system allocates the necessary different resources to handle tasks like message passing, transaction ordering, and state computation. For users, the underlying complex distributed architecture is abstracted, so even a large cluster of nodes behaves like a single computer. This is because the AO system uses SSI to integrate the complex components of distributed systems, achieving a unified computing environment. In other words, through the SSI architecture, AO integrates multiple distributed computing nodes into a unified resource, providing users with a transparent, efficient, scalable, and unified computing platform.
Five, Resource-driven Value Creation and Application Innovation
In summary, through the combination of SCP, Actor, and SSI, AO has built an innovative architecture that provides three scalable system-level resources for computing, storage, and data (consensus), along with a security resource supported by $AO. Resources, as core production factors, play a key role in promoting technological advancement, stimulating application innovation, and improving economic efficiency. By clarifying the resource elements in the AO + Arweave system, we can optimize resource planning and management, leveraging resources to drive technological and application innovation, accelerate value creation in Web3, and promote the growth of the crypto economy.
Here, we provide a summarizing overview:
1. Infrastructure-based Value Creation
Decentralized World Computer: AO integrates scalable computing, storage, and data resources, providing a unified decentralized computing platform for all applications, with verifiable and trust-minimized features. Applications only need to focus on business innovation, avoiding redundant efforts, making AO a public infrastructure for application innovation.
On-chain Shared Data Resource Library: Arweave can permanently store almost all types of data, becoming a 'Library of Alexandria' that never disappears. Whether financial data or non-financial data, its immutable and verifiable characteristics make it a public good that offers consensus value, supporting combinatorial innovation.
Customizable Security Facilities: AO can provide customized security mechanisms for clients and applications based on different data types and values, achieving a balance between security, cost, and efficiency.
Bridge between Web2 and Web3: AO operates off-chain and can seamlessly integrate with on-chain and off-chain systems, becoming a connecting bridge between Web2 and Web3. Any Web2 application can initiate processes in AO through API and messaging mechanisms, calling network units in AO to perform computations while customizing their security mechanisms.
2. Technological and Application Innovations
Blockchain has developed to date, with public chains led by Bitcoin, Ethereum, Solana, etc., whose applications remain biased towards the financial sector, such as asset issuance, trading, mortgage lending, derivatives, etc., leading many to mistakenly believe that the role of blockchain is limited to this.
However, the innovative architecture of AO + Arweave adds new feasibility for technological innovation and application development in blockchain. In addition to supporting financial innovations typical of most public chains, AO, as a general-purpose world computer, supports all data types and corresponding application innovations, especially non-financial data-driven application innovations.
Loading AI Models: The AO + Arweave architecture provides infinite computing, storage, and data resources, enabling AO to run various open-source large language models directly in smart contracts, such as Llama 3 and GPT-2, supported by three key technologies: WASM64, WeaveDrive, and the Llama.cpp large language model inference engine, allowing smart contracts to directly handle complex data and make concurrent decisions, such as the on-chain autonomous virtual world Llama Land powered by the AI-driven Llama 3 model.
Creating Agent and AgentFi: Based on the reasoning capabilities of AI models, as well as the ability of AO processes to respond to implicit messages over time, awaken themselves, and execute actions, along with the capability to 'subscribe' to a process by paying fees to MU, thus triggering calculations at appropriate frequencies, AO supports Agents and AgentFi that can meet complex business logic, predefined requirements, and diverse autonomous strategies.
Copyright Management and Creator Market (ContentFi): Arweave stores various types of data in atomic assets, making data easy to identify and confirm ownership. It can be monetized as a new form of digital asset through circulation and trading in the market, establishing clear benefit distribution and collaboration patterns to support copyright management and the creator market.
Next Generation Internet Framework Permaweb: Unlike the three-layer structure of the application layer, service layer, and storage layer of traditional Web2 internet, Permaweb achieves permanent storage of all content by replacing the storage layer with Arweave's permanent storage solution, storing it in Arweave as atomic assets. It builds applications supporting AO's super-parallel computing based on SCP at the application layer, creating a perpetually online, decentralized next-generation internet framework. This framework integrates with Web2 and provides an experience indistinguishable from Web2, yet there are significant differences; Permaweb is not a 'walled garden.' It offers developers, operators, and users a fair and open environment: users own and control their data; data flows freely between different applications; developers and operators can utilize data for business within established rules without special permissions, thus promoting mutual benefits among all parties.
The above are several typical application innovation directions that AO can support. Of course, AO can support more data types and broader scenarios for application innovation. Although the AO ecosystem has developed for a short time and technological and application innovations require time for testing, we prefer to assess the significance and value of these innovations from the entire Web3 industry's development stages and the characteristics of Web2 systems.
Currently, the Web3 industry is exploring feasible paths for large-scale adoption. Many blockchains are striving for this, such as TON combining with Telegram, guiding real users from Web2 to real applications in Web3, intending to achieve large-scale conversion of traffic to liquidity value; CKB is becoming Bitcoin's L2, constructing a lightning network based on CKB, aimed at bringing high-frequency, small-amount, large-scale peer-to-peer payments.
From an industry development perspective, AO + Arweave redefines the implementation framework of decentralized computers, bringing system flexibility, security, and economic efficiency through innovative architecture, building scalable system-level resources, sustainably releasing resource potential, driving technological and application innovations, realizing value creation and transfer, promoting the integration of Web3 and Web2, and providing a feasible path for Web3 to achieve large-scale adoption.
References
1. A Protocol for Sustainably and Permanently Retaining Information
2. AO Protocol: Decentralized, Permissionless Supercomputer:
https://x.com/kylewmi/status/1802131298724811108
3. Storage-based Computing Paradigm Implemented by Arweave:
https://news.ever.vision/a-storage-based-computation-paradigm-enabled-by-arweave-de799ae8c424
4. Technical Explanation of AO Super Parallel Computer:
https://www.chaincatcher.com/article/2121544
5. Interpretation of SCP: Breaking Away from the Fixed Infrastructure Paradigm of Rollup Trustlessness:
https://mp.weixin.qq.com/s/BPRAsby78G2a835pX1l3iw
6. In-depth Analysis of the Actor Model (Part 1): Introduction to Actor and its Application in the Gaming Industry:
https://blog.csdn.net/weixin_44505163/article/details/121191182
7. Arweave Permanent Storage + AO Super Parallel Computer: Building a Data Consensus Infrastructure:
https://www.chaincatcher.com/article/2141924
