Introduction🔗
The Performance examples measure how fast the 51Degrees device detection engine processes evidence (User-Agents and User-Agent Client Hints) on your own hardware. They run a tight loop over a fixed set of User-Agent strings, count detections per second, and report the mean time per detection, so you can size capacity, validate tuning changes, and compare deployment models on the platform you actually run.
The companion Performance options page describes the configuration knobs that drive throughput and memory: the Performance Profile (LowMemory, Balanced, HighPerformance, MaxPerformance), how the data file is loaded, and the size of the property and value caches. Pick a profile, run the example, and use the numbers to decide whether to trade memory for throughput on your target hosts. The example covers the on-premise Hash engine. Cloud performance depends on network round-trip time and is not benchmarked here.
The example is available in C, C#, Java, Node.js, Python, and Rust. The C, C#, and Java versions process 20,000 sets of HTTP headers, some with User-Agent Client Hints, and the Node.js, Python, and Rust versions process 20,000 User-Agents, so compare runtimes only within each group. For C++, use the C example. PHP has its own benchmark, built on ApacheBench.
Results we measured on seven machines are in the hardware benchmarks.