Lagrange 백서

Lagrange

원문

bitpro가 원문에서 글자만 모은 내용이에요. 표·그림·수식은 빠지거나 깨질 수 있어요. 전체 2쪽 중 1쪽

Whitepaper - Lagrange Prover Network: A Decentralized, Scalable Zero-Knowledge Proving Ecosystem Abstract Lagrange Labs introduces the Lagrange Prover Network, a pioneering framework designed to deliver a modular, infinitely scalable proving layer for zero-knowledge (ZK) proof generation. Central to this ecosystem are Prover Supernets, customizable, isolated subnetworks that enable anyone to deploy specialized proving environments within the broader Lagrange Network. Each Prover Supernet offers unique flexibility in defining tokenomics, staking and hardware requirements, prover permissioning, and auction mechanisms, tailored to specific computational needs. The architecture hinges on the interaction between Gateways, which orchestrate task distribution and define operational rules, and Provers, which supply computational capacity to generate ZK proofs. This model empowers each Supernet to autonomously configure its ecosystem, aligning with strategic goals while capturing value for the Lagrange native token through proving activities. Deployed on EigenLayer and backed by over 85 institutional-grade operators, the Lagrange Prover Network ensures high liveness, cost efficiency, and decentralization. This whitepaper explores its innovative architecture, economic model, and real-world applications - highlighted by case studies of the ZK Coprocessor Supernet and ZKML DeepProve demonstrating how it empowers clients like rollups, ZK coprocessors, ZKML, and decentralized applications (DApps) to access tailored proving resources. By integrating a sustainable tokenomics model, the network positions itself as foundational infrastructure for blockchain scalability and interoperability. Introduction Zero-knowledge proofs are a cornerstone of blockchain technology, enabling enhanced privacy, scalability, and trustless computation. However, the computational intensity of ZK proof generation poses significant challenges, including high operational costs, latency, and scalability limitations in traditional proving systems. These constraints hinder the widespread adoption of ZK-based applications, particularly for large-scale ecosystems like rollups and interoperability protocols. Lagrange Labs addresses these challenges with the Lagrange Prover Network, a decentralized and modular proving infrastructure designed to scale dynamically with demand. Through Prover Supernets - customizable subnetworks within the Lagrange ecosystem—the network delivers a flexible, production-ready solution, abstracting the complexities of distributed proving. This whitepaper delves into the Lagrange Prover Network’s architecture, economic model—including how proving profits bolster the native token—and unique features. Case studies on the ZK Coprocessor Supernet and ZKML illustrate its real-world impact, showing how it empowers developers and protocols to unlock new use cases and elevate Web3 usability. Architecture The Lagrange Prover Network is engineered as a "network of networks," comprising multiple independent subnetworks, or Prover Supernets, each with dedicated proving bandwidth. This modular design eliminates bottlenecks and enables customization of economic models including tokenomics that drive value to the native token, hardware requirements, and operational rules, tailoring each Supernet to specific proving needs. Core Components 1. Gateway ○ Role : Acts as a lightweight mediator between clients (e.g., ZK Coprocessors, Rollup Chains, or Layer 2 solutions) and Provers, orchestrating task distribution and proof aggregation. ○ Functionality : Manages a queue of proving tasks, conducts auctions to assign jobs, and ensures seamless communication via a gRPC server . It defines the Supernet’s economic and operational parameters, such as staking requirements and auction mechanisms. ○ Key Features : Persistent storage for task resilience, automatic job reassignment for fault tolerance, and support for standardized or custom communication schemas (e.g., protobuf). 2. Provers ○ Role : Serve as the computational nodes that generate ZK proofs based on tasks assigned by the Gateway. ○ Functionality : Bid on proving jobs through auctions, execute proof generation using preloaded circuit parameters, and submit results back to the Gateway. ○ Key Features : Application-specific optimization through preloaded parameters, global distribution on bare-metal servers, and adherence to strict Service Level Agreements (SLAs). Each Supernet operates autonomously, enabling specialization for high-throughput environments while maintaining isolation from other subnetworks. This architecture ensures scalability and security, as performance issues in one Supernet do not affect others. Workflow The workflow of the Lagrange Prover Network integrates a robust auction mechanism to allocate proving tasks efficiently, ensuring competitive pricing and high reliability. Below is a detailed breakdown of the process, incorporating insights from the Gateway lifecycle, Prover operations, and the auction design. Gateway The Gateway orchestrates the entire proving process, from task submission to proof delivery, with auctions playing a pivotal role in job assignment. 1. Task Submission ○ Clients submit computational tasks requiring ZK proofs to the Gateway via the gRPC interface. ○ Tasks are logged in a persistent Storage Layer , ensuring recoverability if connections are disrupted. 2. Prover Onboarding and Availability ○ Provers join the Supernet by staking required tokens (e.g., ETH via EigenLayer or Supernet-native tokens) and meeting hardware specifications (e.g., multi-core CPUs or GPUs). ○ In permissioned Supernets like the ZK Coprocessor, Provers must be pre-approved, enhancing security against Sybil attacks. ○ Provers announce their availability to the Gateway, specifying supported job categories (small, medium, large) based on their computational capacity. 3. Job Creation ○ The Gateway’s Dispatcher decomposes each task into smaller jobs or batches of jobs, enabling parallel processing across multiple Provers. ○ Batching, as used in the ZK Coprocessor Supernet, reduces auction overhead and allows Provers to bid on preferred proof types. 4. Auction Initiation ○ The Gateway announces jobs or batches for auction, inviting bids from eligible Provers. ○ In the ZK Coprocessor Supernet, batches are announced every few seconds, optimizing throughput and minimizing latency. 5. Bidding Phase ○ Provers submit sealed bids via gRPC, specifying the price they are willing to accept for completing the job or batch. ○ The auction mechanism is a first-price sealed-bid reverse auction : the Prover offering the lowest bid wins and is paid their bid amount upon completion. This contrasts with traditional FPSB auctions (highest bid wins) and aligns with service procurement, where cost efficiency is prioritized. The auction mechanism is described in more detail below. ○ Confidentiality of bids encourages Provers to bid their true costs, balancing profitability and competitiveness. 6. Winner Selection and Job Assignment ○ The Gateway selects the lowest bidder for each job or batch and assigns the work to the winning Prover. ○ Assignment details are communicated via gRPC, locking the Prover into a deadline-bound commitment. 7. Proof Generation ○ The assigned Prover generates the ZK proof using preloaded circuit parameters, downloaded at startup (often gigabytes of data), to avoid per-task latency. ○ Specialization ensures rapid execution tailored to the Supernet’s proof types (e.g., rollup validity proofs or coprocessor queries). 8. Proof Submission and Verification ○ The Prover submits the completed proof to the Gateway, where it is stored in the database and verified for correctness. ○ Upon validation, the job is marked as complete, and the Prover is paid their bid amount. 9. Aggregation and Delivery ○ Once all jobs for a task are completed, the Gateway aggregates the proofs into a final output. ○ The aggregated proof is returned to the client via the bidirectional gRPC channel, fulfilling the request. 10. Fault Handling ○ If a Prover misses the deadline, the Gateway re-auctions the job to another Prover, enforcing liveness through staking penalties (e.g., slashing) or non-payment. ○ This ensures continuous progress without client intervention. Provers Provers execute the computational heavy lifting, optimized for efficiency and reliability within the auction-driven framework. 1. Initialization ○ At startup, Provers preload application-specific circuit parameters into memory, enabling rapid proof generation without repeated downloads. ○ This one-time setup contrasts with containerized systems, reducing latency significantly. 2. Availability Announcement ○ Provers signal readiness to the Gateway, detailing their hardware capabilities and supported job categories (small, medium, large). ○ This self-classification ensures tasks align with their strengths. 3. Bidding and Job Execution ○ Provers participate in auctions, submitting bids for jobs or batches based on their cost structures and capacity. ○ Upon winning, they generate proofs using preloaded parameters and submit them before the deadline. 4. Liveness and Penalties ○ Provers must adhere to SLAs, with penalties (e.g., staking slashes or nonpayment) for late or failed submissions, reinforcing network reliability. DARA - Auction Mechanism The auction mechanism, developed by Lagrange, is a cornerstone of the Lagrange Prover Network, particularly in the ZK Coprocessor Supernet, where it is implemented as a first-price sealed-bid reverse auction . Key aspects include: ● Design : Provers bid the price they are willing to accept for proof generation, and the lowest bidder wins, receiving their bid amount upon successful completion. This reverse structure incentivizes cost efficiency, benefiting clients by minimizing fees. ● Batching : In the ZK Coprocessor Supernet, tasks are auctioned in batches every few seconds, reducing auction frequency and allowing Provers to express preferences for specific proof types (e.g., small vs. large jobs). ● Sealed Bids : Confidentiality prevents Provers from colluding or gaming the system, promoting fair competition. ● Permissioned Mitigation : The permissioned model restricts participation to vetted Provers, preventing Sybil attacks where fake identities could manipulate bids. ● Flexibility : Other Supernets can adopt alternative Transaction Fee Market (TFM) mechanisms (e.g., second-price auctions or stake-based assignment) tailored to their goals, though the FPSB reverse auction balances simplicity and efficiency. Persistence and Fault Tolerance The Gateway ensures resilience through: ● Persistent Storage : Task metadata and proofs are stored durably, allowing clients to retrieve results after reconnection without resubmission. ● Automatic Reassignment : Jobs timing out are re-auctioned, maintaining workflow continuity. ● Guaranteed Outcomes : Every task results in a successful proof or a clear error, unlike one-shot systems requiring client retries. Efficiency and Decentralization Benefits ● Specialization : Preloaded parameters and job categorization enable Provers to execute tasks with minimal overhead, ideal for high-throughput applications. ● Competitive Pricing : Auctions ensure cost-efficient allocation, with Provers incentivized to bid competitively. ● Global Distribution : Provers on bare-metal servers worldwide enhance censorship resistance and reduce reliance on centralized providers, bolstering security and uptime. The Lagrange Prover Network’s workflow and architecture combine a decentralized prover pool, a sophisticated Gateway orchestrator, and a reverse auction mechanism to deliver scalable, reliable ZK proof generation. By integrating competitive bidding with fault-tolerant task management, the network optimizes resource allocation, minimizes costs, and ensures liveness, making it a robust foundation for blockchain applications requiring high-performance proving. Case Study: ZK Coprocessor The ZK Coprocessor serves as a flagship example of the Lagrange Prover Network’s Prover Supernet architecture, demonstrating its ability to deliver tailored, high-performance proving solutions. Launched on the mainnet in June 2024, this Supernet is powered by over 60 independent provers, including industry leaders such as OKX, Figment, P2P, and Kraken. It showcases how the network’s flexibility and scalability can support complex, high-throughput ZK applications. Key Features and Customizations ● Hyper-Parallel Processing : The ZK Coprocessor Supernet employs a hyper-parallel approach, breaking down intricate proving tasks into smaller subtasks distributed across multiple provers. This enables horizontal scaling of proving time, efficiently processing large datasets and supporting demanding applications like real-time data queries or rollup computations. ● Hardware Specifications : Provers must operate medium CPU instances with 20–40 vCPUs and 40–80 GB of RAM. This requirement optimizes performance for the Supernet’s proof types, favoring multiple smaller instances over single large ones to maximize parallelism. ● Transaction Fee Market (TFM) : A first-price sealed-bid reverse auction governs task allocation. Provers submit confidential bids, and the lowest bidder wins, receiving their bid amount upon completion. Batched auctions, conducted every few seconds, reduce overhead and allow provers to target preferred proof categories, enhancing cost efficiency and throughput. ● Permissioned Access : Only vetted provers are admitted, bolstering security and reliability by mitigating risks like Sybil attacks, while ensuring a trusted operator pool. ● I/O Integration : The Supernet incorporates prebuilt modules—Scraper for on-chain data requests and Query Executor for on-chain proof submission—streamlining blockchain interactions and reducing latency. ● Staking Commitment : Provers stake 20–60 ETH via EigenLayer’s restaking mechanism, aligning their incentives with network liveness and performance. This stake acts as collateral, enforceable through penalties for missed deadlines. ● Payment Model : Initially, fees are paid in ETH, with a planned transition to Lagrange’s native token, reinforcing the network’s economic ecosystem. Case Study: DeepProve - zkML DeepProve is a groundbreaking zero-knowledge machine learning (zkML) library that enables fast, scalable, and veriable AI inferences, achieving speeds up to 158x faster than leading zkML solutions. By utilizing zero-knowledge proofs (ZKPs), DeepProve allows developers to cryptographically prove that AI models execute correctly and produce intended outputs, all without disclosing underlying data or model specifics. This verifiability is vital as AI increasingly influences critical sectors like healthcare, finance, defense technologies, and governance, ensuring trust and alignment with human interests. Within the Lagrange Prover Network, DeepProve can be deployed as a specialized Prover Supernet, mirroring the adaptability of the ZK Coprocessor. This integration leverages the network’s decentralized, scalable infrastructure to meet the unique computational demands of zkML proof generation. Customization and Flexibility As a Prover Supernet, DeepProve can tailor its operational framework, including: ● Hardware Requirements : Optimized for AI inference workloads, the Supernet can mandate high-performance hardware like GPUs or TPUs to accelerate proof generation for complex machine learning models. ● Staking Mechanisms : Provers may stake Lagrange’s native token or other designated assets, ensuring commitment to network performance and liveness. ● Auction Mechanisms : A customized Transaction Fee Market (TFM), such as a first-price sealed-bid auction, can efficiently distribute proving tasks, balancing cost and speed for zkML computations. This adaptability ensures the DeepProve Supernet is finely tuned to support verifiable AI inferences at scale. Benefits of Integration Deploying DeepProve on the Lagrange Prover Network unlocks several advantages: ● Scalability : The network’s modular design enables the Supernet to expand horizontally, handling rising demand for zkML proofs without performance trade-offs. ● Decentralization : A global pool of provers enhances security and resilience, minimizing dependence on centralized systems. ● Cost Efficiency : Competitive auctions and protocol emissions subsidize proving costs, making zkML accessible to a broader range of developers and organizations. As AI adoption accelerates, the demand for verifiable computations will surge. Integrating DeepProve with the Lagrange Prover Network positions zkML as a transformative use case for decentralized proving, driving innovation across industries. This collaboration not only scales AI’s potential but also ensures its outputs remain trustworthy and reliably serve humanity’s best interests. Tokenomics: Value Capture to the Lagrange Native Token The Lagrange Prover Network’s tokenomics are designed to establish a self-sustaining ecosystem where the native token captures value directly from decentralized proving activities. By integrating a work-based incentive model with flexible staking and emission mechanisms, the network aligns the interests of clients, provers, and token holders. This section outlines how profits from proof generation flow to the Lagrange native token, ensuring economic stability, scalability, and long-term value accrual for the ecosystem from subsidized proving costs. Overview of the Economic Model The Lagrange Prover Network balances token suppl
프로젝트로 돌아가기