An empirical performance and memory benchmarking study evaluating leading open-source ERC-4337 bundler implementations under heavy PQC loads.
Analyzing the architecture, language design choices, and concurrency models of leading bundlers: Alto (TypeScript/Pimlico), Skandha (TypeScript/Etherspot), and Rundler (Rust/Alchemy).
We subject each bundler to simulated bursts of 10,000 post-quantum UserOperations per minute. We evaluate CPU utilization, Node.js event-loop bottlenecks versus Rust multi-threaded async executors, and memory footprint stability under large calldata payloads.
Tuning recommendations for production bundler operators: memory allocation limits, garbage collection parameters, RPC rate-limiting, and distributed Redis queue clustering.