Groundhog: Linearly-Scalable Smart Contracting via Commutative Transaction Semantics

TL;DR

Groundhog introduces a scalable smart contract execution engine that allows concurrent transaction processing using commutative semantics, significantly increasing throughput without contention.

Contribution

It proposes a novel commutative semantics framework and a reserve-commit process to enable high-throughput, conflict-free concurrent execution of smart contracts.

Findings

Processes over 500,000 transactions per second on 96 CPU cores.

Supports a wide range of applications with flexible semantics.

Achieves high scalability without transaction contention.

Abstract

Groundhog is a novel design for a smart contract execution engine based around concurrent execution of blocks of transactions. Unlike prior work, transactions within a block in Groundhog are not ordered relative to one another. Instead, our key design insights are first, to design a set of commutative semantics that lets the Groundhog runtime deterministically resolve concurrent accesses to shared data. Second, some storage accesses (such as withdrawing money from an account) conflict irresolvably; Groundhog therefore enforces validity constraints on persistent storage accesses via a reserve-commit process. These two ideas give Groundhog a set of semantics that, while not as powerful as traditional sequential semantics, are flexible enough to implement a wide variety of important applications, and are strictly more powerful than the semantics used in some production blockchains today. Unlike prior smart contract systems, transactions throughput never suffers from contention between transactions. Using 96 CPU cores, Groundhog can process more than half a million payment transactions per second, whether between 10M accounts or just 2.