From Synthetic Benchmarks to Feedback-Controlled Testbeds: Observability and Workload Management in a Containerised Solana Localnet

This archive contains the LaTeX source, generated PDF, bibliography, workload-generation script, and patch evidence for the article "From Synthetic Benchmarks to Feedback-Controlled Testbeds: Observability and Workload Management in a Containerised Solana Localnet". The article argues that synthetic benchmarking in distributed systems is evolving from open-loop performance measurement towards reproducible experimental testbeds that combine workload generation, observability, dataset production, and future feedback-control interfaces. The case study is a containerised private Solana localnet with validator, wallet/workload components, Yellowstone/Geyser-based observability, monitoring services, and Prometheus-compatible metrics. This version explicitly documents the transaction-load mechanism used in the empirical pilot. The controlled workload profiles `txload-low` and `txload-medium` were implemented using `scripts/run_article_txload_container_v0_4.sh`. The script creates temporary payer and recipient key pairs within the validator container, funds the payer account via a local airdrop, and then sequentially submits native SOL transfer transactions via the local RPC endpoint. Each transfer sends 0.000001 SOL from the payer to the recipient and records workload-side evidence, including timestamps, sequence numbers, exit codes, durations, signatures, per-transfer logs, and a summary JSON. The empirical pilot is intentionally limited. It validates the measurement pipeline and the distinguishability of controlled transaction-load profiles. It should not be interpreted as a general Solana mainnet performance benchmark, a multi-validator scalability study, or a production-like application-level workload evaluation.