Optimizing Blockchain Transaction Execution Performance by Applying Genetic Sequencing

Abstract

The successive generations of consensus algorithms progressively shifted the performance bottleneck of blockchains to the execution layer. Recent works have addressed this bottleneck by parallelizing the execution of transactions. Historically, transaction ordering was left to the discretion of validators, a practice that lacked transparency and gave rise to Maximal Extractable Value (MEV) attacks where transaction ordering is manipulated for private gain. More recently, the focus has shifted toward fair ordering protocols that prioritize chronological submission. However, fair ordering is often misaligned with validator incentives and negatively impacts execution throughput under high congestion. In this work, we address the tension between validator revenue and fair ordering using a dynamic optimization framework.

We define a blockchain-independent model to evaluate transaction ordering in a continuous setting where the execution of successive blocks can overlap. Within this model, we propose an anytime genetic algorithm. We use real-world blockchain data and execution time estimates within realistic error margins, showing that this approach increases validator profit by around 15% and accelerates congestion relief. We also quantify the impact of adding fair ordering constraints on validator revenue during congestion, showing that revenue decreases by around 50%.