Access Specification-Aware Software Transactional Memory Techniques for Efficient Execution of Smart Contract Transactions

2025-03-05

AI Summary

To address high transaction abort rates and limited throughput in blockchain smart contract parallel execution using Software Transactional Memory (STM), this paper proposes saSupraSTM—the first STM design that deeply integrates user-provided read-write access specifications into its core algorithm. Our approach leverages specification-driven conflict prediction, scheduling optimization, and spurious abort suppression, combined with access-aware STM primitives, MoveVM/EVM-compatible schedulers, a multicore parallel execution engine, and lightweight static/dynamic read-write set inference. Evaluated on both EVM and MoveVM, saSupraSTM reduces abort rates by 37% and improves peak throughput by up to 2.1× over specification-aware variants of BlockSTM and PEVM. These results establish a new performance benchmark for on-chain transaction execution.

Technology Category

• Data Mining & Knowledge Management: Scalability, Parallel & Distributed Systems
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories
• Multiagent Systems: Mechanism Design

Application Category

• Security and Privacy: Cryptocurrency and smart contracts
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization

Abstract

For a high-performance blockchain like Supra's Layer 1, minimizing latencies across key components is crucial-such as data dissemination, consensus (or ordering), and transaction execution. While through significant innovations we have improved the first two, transaction execution remains an area for further optimization. Software Transactional Memory (STM) is a widely used technique for parallel execution, with Aptos' BlockSTM pioneering its application of efficient blockchain transaction processing on multi-core validator nodes. Subsequently, PEVM [13] adapted BlockSTM for EVM transaction execution. However, we identified a gap in existing STM techniques-while access specifications have been used in industry (e.g., Solana's user-provided read-write sets), they have not been leveraged to enhance STM efficiency. Our experimental analysis demonstrates that specification-aware STMs outperform their plain counterparts on both EVM and MoveVM. To maximize these benefits, we have designed specification-aware SupraSTM (saSupraSTM), a novel algorithm that fully utilizes access specifications. Through extensive testing, saSupraSTM outperforms both our specification-aware adaptation of Aptos' BlockSTM and specification-aware PEVM, setting a new benchmark for transaction execution efficiency in the context of blockchain networks.

Problem

• Optimize transaction execution in high-performance blockchains.
• Enhance Software Transactional Memory (STM) using access specifications.
• Develop saSupraSTM for superior blockchain transaction efficiency.

Innovation

• Leverages access specifications for STM efficiency
• Introduces saSupraSTM for optimized transaction execution
• Outperforms existing STM techniques on EVM and MoveVM