Actual performance test data for scaling solutions varies by system and workload. For cloud-based horizontal scaling (e.g., AWS EC2), adding nodes linearly boosts throughput—doubling instances often increases request handling by ~80-90% under 10,000 concurrent users, with latency rising <10ms. Vertical scaling (e.g., upgrading CPU/RAM) shows diminishing returns; a 2x CPU boost may yield only 30-50% performance gain due to bottlenecks like I/O. Kubernetes auto-scaling maintains <5% error rates up to 50,000 requests/second, but cold starts add 100-500ms latency. Database sharding (e.g., MongoDB) improves read/write speeds by 60-70% per shard, though cross-shard queries spike latency by 20-30ms. Real-world tests (e.g., e-commerce platforms) show 99.9% uptime during 100,000-user spikes with hybrid scaling, but costs rise 2-3x. Always validate with your workload—benchmarks differ.