Network-Controlled Dataset of Blockchain Benchmarks for Evaluating Repeatability and Performance Predictability

This release archives the datasets, analysis code, frozen topology and workload inputs, validation tests, and generated evidence used for controlled, topology-aware blockchain benchmarking and for the run-to-run repeatability analysis in the IEEE Access manuscript Experimental Repeatability and Performance Predictability in Network-Controlled Blockchain Evaluation. The raw repeated-execution dataset contains 4,105 observations. The documented fixed selection yields 4,080 independently reset executions across 300 blockchain-topology-workload-size configurations, covering five blockchains, five network topologies, six workloads, and validator sets of 10 and 40 nodes. The selected corpus contains 3,125 positive-commit executions, 953 zero-commit executions, and two no-submission executions. The archive includes explicit outcome and metric-eligibility rules; configuration-level dispersion and tail analyses; a balanced nine-run sensitivity; 5,000-repetition within-configuration bootstrap intervals; two-part service-incidence and conditional-performance models; targeted three-way sensitivities; within-blockchain REML random-intercept ICCs with configuration-cluster bootstrap intervals; topology and campaign audits; manuscript-ready tables and figures; and a 65-test validation suite. This version extends the previous Zenodo release with the complete revision workflow and its frozen, auditable outputs.