Poresi Whitepaper

Mar 3, 2025

1. Abstract

Poresi introduces a VM-agnostic Execution Framework that optimizes blockchain execution through a modular design. Its three-module architecture enables high-performance execution, multi-chain interoperability, and improved scalability. Additionally, Poresi evolves into a foundational infrastructure for AI-driven on-chain applications, offering an execution environment tailored to the demands of decentralized AI computation.

The Poresi Stack decouples blockchain execution layers from single binary implementations and network design, enabling seamless integration with any Layer 1s, Layer 2s, and application-specific chains for instant performance boosts and enhanced interoperability. The architecture aims to achieve Web2 parity for on-chain use cases in a multi-chain future, ensuring onboarded chains are futureproofed for performance and AI compatibility.

2. Introduction

Problem Statement

Blockchain systems today face significant challenges, including network congestion, high transaction costs, and limited scalability for opinionated use cases—even as generalized blockspace continues to expand. These limitations stem from monolithic execution layers and insufficient interoperability, which put networks at risk of becoming obsolete without scalability solutions like parallel execution.

Meanwhile, the rise of decentralized AI agents, on-chain model inference, and autonomous data economies has created new requirements for execution environments that can support high-frequency, parallelizable AI tasks. Many chains lack the specialized infrastructure to meet the performance demands of AI computation.

Poresi addresses these challenges by offering a turnkey modular execution layer that supports both general-purpose and AI-native workloads. This enables chains to scale efficiently, futureproof their infrastructure, and onboard AI-driven applications without building bespoke systems from scratch.

Vision

Poresi’s mission is to enhance blockchain scalability and interoperability at the execution layer, enabling applications and ecosystems—including AI applications—to achieve their full potential. Poresi aims to empower multi-chain ecosystems with shared execution infrastructure focused on:

Efficient scaling solutions with continuous upgrades that maintain optimal performance

Cross-chain interoperability

Developer incentives for building applications designed for multi-threaded, parallel processing environments

AI-native execution support for model inference, agent coordination, and decentralized intelligence tasks

Scope

This paper covers Poresi's architecture, implementation, and development roadmap, with an emphasis on its role as a high-performance infrastructure for AI-powered blockchain applications.

3. Technical Architecture

Poresi's technical architecture uses three core pillars to optimize blockchain execution:

  1. Parallel Scalable Storage: Enables concurrent operations and efficient state updates across nodes.
  2. Instruction-level Parallelism: Uses Static Single Assignment (SSA) to identify and re-execute conflicting instructions independently.
  3. Application Code Assessor: Rewards developers for creating optimized smart contracts in execution environments.

Together, these pillars enhance scalability, efficiency, and interoperability in blockchain execution.

Parallel Scalable Storage

State management is crucial for blockchain performance, as each transaction updates the system state. Traditional systems using Merkle-Patricia tries on SSDs face latency and I/O bottlenecks, which worsen as blockchain activity grows. Parallel Scalable Storage offers a new approach to state data management.

Poresi's Parallel Scalable Storage solution uses sharded state management, distributed caching, and advanced data structures to achieve high performance with low latency. This article explores its key features and benefits.

Core Principles of Parallel Scalable Storage

The system partitions state data into independently managed shards across distributed nodes, enabling parallel operations. Shards are distributed by key prefixes for balanced workloads. The Scalable State Merkle Trie (SSMT) implements this by assigning specific prefix ranges (e.g., 0x00-0x10) to different nodes.

To minimize latency, distributed caching keeps frequently accessed data in memory, using prefetching and historical patterns. The SSMT optimizes state management through efficient query handling and workload balancing algorithms, preventing performance bottlenecks.

Architectural Innovations

Parallel Scalable Storage features innovative dynamic load balancing for even shard distribution, concurrency-optimized algorithms for parallel processing, memory-first design for faster access, and fault tolerance for reliability.

SSMT Evaluation and Benchmarks

We evaluated SSMT's performance using the Ethereum mainnet dataset, comparing transactions-per-second (TPS) and gas-per-second (GPS) metrics of Poresi's execution engine with Deterministic OCC and SSMT. Testing on a 32-core 128GB server with 100GB high-performance SSD showed 50% reduced I/O overhead and 1.2-1.5x performance improvement for large datasets.

SSMT integration increased average GPS to ~657M, achieving 1.8x improvement over Deterministic OCC alone and 30x over regular GPS on Ethereum mainnet.

Combined with SSMT, the average TPS increased to ~6635, which is 1.6x of TPS offered by Deterministic OCC alone and 25x of the regular TPS, based on the Ethereum mainnet data.

Parallel Scalable Storage uses distributed memory-first storage, unlike traditional disk-based blockchain systems. While other high-performance designs rely on expensive high-capacity servers, Poresi achieves better efficiency through caching, sharding, and load balancing on standard hardware. Instead of centralized SSD-based storage that causes I/O bottlenecks, our system distributes shards with in-memory caching for faster access.

More importantly, beyond reducing hardware costs, Poresi is future-proving—it scales efficiently with distributed storage, while other high-performance designs rely on single-machine storage that limits scalability as demand grows.

Conclusion

Parallel Scalable Storage represents a paradigm shift in blockchain state management. By combining sharded architecture, distributed caching, and optimized data structures such as the Scalable State Merkle Trie (SSMT), it overcomes the limitations of traditional systems, offering unparalleled throughput, scalability, and cost efficiency. As blockchain networks evolve, the adoption of such advanced solutions will be essential to meet the demands of a growing decentralized ecosystem.

Instruction-level Parallelism

Traditional blockchain systems use sequential transaction execution, creating bottlenecks at scale due to complete transaction re-execution during conflicts.

Poresi's Instruction-Level Parallelism (ILP) processes individual instructions concurrently, reducing conflicts and improving performance by identifying dependencies between granular instructions.

Core Principles of Instruction-Level Parallelism

Static Single Assignment (SSA) transforms smart contract bytecode to track instruction dependencies, enabling concurrent execution. Key principles:

  • Granular Execution: Process independent instructions simultaneously
  • Dependency Graph Analysis: Manage instruction relationships for optimal sequencing
  • Conflict Minimization: Reduce re-execution needs through instruction-level tracking

Key ILP optimizations include:

  • SSA Transformation: Converts bytecode to track dependencies efficiently
  • Dynamic Scheduling: Parallel processing based on dependencies
  • Incremental Execution: Continuous progress despite conflicts

ILP enhances throughput, reduces latency, and maximizes resource usage while providing developers with optimization tools.

SSA Evaluation and Benchmarks

SSA Evaluation uses dependency graphs to identify conflict-causing instructions during execution. Performance is measured by comparing critical path length to original Gas consumption, determining SSA efficiency gains.

Poresi achieves 75% gas cost reduction through selective instruction re-execution.

Conclusion

Instruction-Level Parallelism transforms blockchain execution through concurrent processing and enhanced efficiency. Using advanced techniques like SSA and Deterministic OCC, ILP improves transaction processing beyond sequential limitations. This innovation will help blockchain networks scale effectively, offering better performance and user experience.

Application Code Assessor

The Application Code Assessor module incentivizes developers to create smart contracts that maximize parallel execution potential.

Deterministic Optimistic Concurrency Control (OCC)

Poresi's Deterministic OCC sorts transactions by assuming parallel execution. When state conflicts occur, transactions are deterministically aborted and re-executed, identifying conflicts while preserving gas efficiency for operations after the conflict point.

During conflicts, the later transaction aborts and re-executes from the conflict point. Multiple aborts waste gas and indicate reduced parallelism.

Evaluation Framework

Poresi implements an evaluation framework to analyze smart contract performance and encourage parallel execution. It measures transaction efficiency and rewards high parallelism.

The system evaluates contracts periodically, analyzing transaction dependencies and calculating efficient gas ratios.

Smart contract parallel execution suitability is evaluated by:

  • Transaction Parallel Degree: Higher throughput indicates better parallelism
  • Total Gas Amount: Higher gas consumption indicates better contract utilization

Contract data is captured at period end for reward calculations.

Deterministic OCC Evaluation and Benchmarks

We tested Deterministic OCC's speedups, TPS, and GPS across thread counts using conflict-free data and Ethereum mainnet measurements.

Deterministic OCC's main contribution is Poresi' Application Code Assessor, which incentivizes reduced transaction conflicts for better performance.

Using a 32-thread server, non-conflicted transactions achieve ~400x speedup over sequential execution.

GPS lifts significantly by adding thread counts, reaching ~70 with a 32-thread domestic server based on non-conflicted transactions.

TPS lifts significantly by adding thread counts, reaching ~2M with a 32-thread domestic server based on non-conflicted transactions.

Deterministic OCC guarantees an average ～7x speedup based on the Ethereum mainnet data.

Deterministic OCC guarantees an average Gas-Per-Second (GPS) of ~374M, which is 16x of the regular GPS, based on the Ethereum mainnet data.

Deterministic OCC guarantees an average Transaction-Per-Second (TPS) of ~4190, which is 15x of the regular TPS, based on the Ethereum mainnet data.

Conclusion

The Application Code Assessor (ACA) is designed to further enhance overall execution performance by encouraging application developers to write more efficient code. It offers tools and protocol-level evaluations that help smart contract developers create applications optimized for a multi-threaded, parallel processing execution environment.

4. Implementation

The Poresi Stack is our implementation of the VM-agnostic Modular Execution Framework and utilizes parallel execution and advanced state management to solve core blockchain inefficiencies.

‍

By separating execution from consensus and leveraging modular data availability layer, the Poresi Stack creates a more flexible and resilient blockchain architecture than traditional solutions like Cosmos SDK or Geth.

Poresi Execution Core

The Poresi Execution Core powers the Poresi Stack through its Parallel Execution Engine and Scalable Data Cache, enabling fast transactions and state management in concurrent operations.

Parallel Execution Engine

The Parallel Processing Engine is the core component of the Poresi Stack, enabling parallelism transactions for higher throughput and lower latency. It leverages modern hardware to handle increasing workloads efficiently.

The engine has three key components:

  • Deterministic OCC (Optimistic Concurrency Control with Deterministic Aborts): Enables parallel transaction execution with consistent conflict resolution.
  • SSA (Static Single Assignment): Tracks instruction dependencies to minimize re-execution overhead.
  • SSMT (Scalable State Merkle Trie): Manages state data using a sharded Merkle tree structure. Data is distributed across nodes by key prefixes, with workload balancing for optimal performance.

Parallel Processing Capabilities for Layer 1s and Layer 2s

The Poresi Stack enables parallel processing in both L1 and L2 through lightweight modifications and customizable solutions.

L1 Integration

Poresi Core integrates with L1 in two ways:

  • L1 Without Decoupling Consensus and Execution: The L1 proposer outputs executed blocks. Poresi Core helps generate these blocks efficiently through parallel execution, removing performance bottlenecks. Examples: Ethereum 1.0 and most of the older L1 designs.
  • L1 with Decoupled Consensus and Execution: The consensus layer proposer outputs unexecuted blocks, which execution nodes then process. Poresi Core enables parallel execution of these blocks for higher throughput. Examples: Ethereum 2.0, some cosmos-based chains such as Berachain.

For decoupled architectures, Poresi Core's deterministic parallel execution ensures consistent results across all nodes, maintaining L1's distributed consistency.

L2 Integration

  • Customizable Settlement Layer: Choose between Ethereum Settlement, Sovereign Settlement, or custom options.
  • Customizable Data Availability Layer: Select from existing data availability providers such as Celestia DA, Eigen DA, or custom solutions.

5. Open Execution Network: Scalable, Permissionless Parallel Execution

Shared Execution System

Poresi introduces a shared execution system designed to address the growing demand for scalable, high-performance transaction processing across diverse blockchain ecosystems. The core of this system is the ability to provide parallel transaction execution to any chain—whether Layer 1s or Layer 2s—on a pay-for-use basis. By offering execution as a service, Poresi enables chains to offload transaction processing, thereby increasing throughput and reducing costs. This infrastructure is also ideal for AI agents and decentralized inference networks, where compute-heavy tasks benefit from Poresi’s concurrency and workload distribution.

Open Execution Network

While the shared execution model allows Poresi to serve as a centralized provider of parallel execution infrastructure, the transition to an Open Execution Network (OEN) marks a key milestone in decentralizing execution and enabling seamless participation from any entity in the blockchain ecosystem.

The Open Execution Network The Open Execution Network aims to provide an open-access platform where any participant—including AI inference providers, decentralized compute networks, or autonomous agents—can leverage parallel execution capabilities without requiring permission or centralized coordination. By moving from a shared execution service to a fully decentralized, permissionless model, Poresi is positioning itself as a backbone for distributed AI computation and high-throughput blockchain operations.

Key enhancements in the OEN model include:

  • Decentralized AI Execution Infrastructure: Executors can run AI models, inference tasks, or data processing jobs in parallel, powered by the Poresi execution engine.
  • Dual-staking for AI Security: Executors may stake Poresi tokens and/or compute-task-specific tokens (e.g., from AI protocols) to align incentives.
  • Cross-chain AI Task Interoperability: AI workflows involving multiple chains or inference outputs from different environments can be processed seamlessly through the OEN.

This framework transforms Poresi from a general-purpose execution system into a foundational layer for AI-native and compute-intensive blockchain use cases.

6. Roadmap

Poresi is set to become the leading solution to enhance blockchain scalability and interoperability at the execution layer via the Poresi Stack. The roadmap highlights key milestones in expanding Poresi’ capabilities across blockchain ecosystems in a multi-chain future.

Timeline:

  • Poresi Execution Engine Core supporting EVM L1/L2 Integration by Q1 2025
  • Expand compatibility to multiple VMs by Q2 2025
  • Shared Execution Layer Launch by Q4 2025
  • Open Execution Network (OEN) Transition by Q4 2026

The Poresi Stack

By decoupling execution from the network design, the Poresi Stack allows seamless integration with L1s, L2s, and application-specific chains, enhancing interoperability across various blockchain solutions. This modular approach supports diverse VM environments, from EVM to MoveVM, CosmWasm, and more, and further provides developers with templates for building optimized applications on top of any chain.

Poresi Stack's flexibility makes it compatible with both existing and newly developed chains, enabling easy upgrades and configurations for blockchain ecosystems.

Shared Execution Layer

Poresi will horizontally integrate the network utilizing the Poresi Stack into a Shared Execution Layer, providing an execution infrastructure that enhances parallel processing across multiple chains. This solution will offer optimized execution resources via dynamic balancing, enabling developers to improve scalability and reduce latency for decentralized applications (dApps) and smart contracts. The Shared Execution Layer will serve as a foundation for broader use cases, facilitating cross-chain operations with improved throughput and efficiency.

Decentralize Further with the Open Executor Network
Poresi plans to transition to a fully decentralized execution model with the Open Execution Network (OEN). This shift will allow any participant, whether chains, applications, or decentralized entities, to contribute computational resources and execute transactions across any blockchain. Executors will stake both Poresi’ token and the native tokens of the chains they serve, aligning incentives and ensuring the network's reliability. This transition will eliminate centralized points of failure and create a permissionless, resilient, and scalable execution ecosystem, enhancing cross-chain interoperability and supporting the multi-chain future of blockchain technology.

7. Conclusion

Poresi is creating VM-agnostic, high-performance execution layers compatible with any chains to futureproof blockchain infrastructure. By integrating AI-native execution support and incentivizing the development of parallelized, intelligent applications, Poresi positions itself at the convergence of blockchain scalability and AI computation.

With Poresi, blockchain ecosystems will have continuous access to industry-leading execution performance, enabling Web2-level efficiency for decentralized apps, smart contracts, and AI agents alike—all while preserving the principles of decentralization.

As AI workloads become increasingly integral to decentralized applications, Poresi will serve as the execution layer of choice for developers building the next generation of on-chain intelligence.

‍

References

Glossary

  • Static Single Assignment (SSA): A compiler optimization technique ensuring each variable is assigned only once. In Poresi, SSA enables instruction-level parallelism (ILP) by tracking dependencies and reducing redundant re-execution.
  • Deterministic Optimistic Concurrency Control (Deterministic OCC): A parallel execution model where transactions initially assume no conflicts. If conflicts arise, affected transactions are deterministically aborted and selectively re-executed, minimizing gas wastage and maximizing throughput.
  • Scalable State Merkle Trie (SSMT): A distributed state storage structure that partitions data across nodes using sharded prefixes. SSMT enhances query efficiency, load balancing, and state synchronization, significantly reducing I/O overhead and improving execution speed.