Teranode Global-Scale Replicated Validator Dataset (1–100 Nodes, 1.0112B TPS Single-Node Aggregate, 1000-Second Runs)

This dataset contains full raw experimental exports for a replicated, partitioned validator architecture evaluated at scales M = 1, 5, 10, 50, and 100 nodes over 1000-second runs. A single node aggregates 919 validator replicas, each sustaining 1.10 × 10⁶ TPS, yielding an observed steady-state aggregate throughput of approximately 1.0112 × 10⁹ transactions per second. Multi-node experiments evaluate WAN-coordinated scaling across ten global regions. Aggregate cluster throughput follows: TPS_cluster = M × R × λ₁ × η(M) where: λ₁ = 1.10 × 10⁶ TPS per validator replica R = 919 replicas per node η(M) = scale-dependent coordination efficiency The dataset includes: Per-second Prometheus metrics (TPS, six latency quantiles per pipeline stage, CPU, memory, disk, network, GC pause, internal queue depth) Aerospike namespace statistics (ops/sec, device utilisation, write amplification) Kafka JMX exports (produce/consume latency, under-replicated partitions, ISR events, controller logs) Per-node time series for every node at each scale Metadata encoding the replica aggregation model Load generation configuration and fault injection parameters Fault injection (Kafka broker throttle, t=650–780s) is included in all runs. The dataset is structured to make explicit the distinction between per-replica throughput (λ₁) and aggregate node throughput.