DeviceAtlas is a best-in-class, high performance, device detection solution. We are fully transparent on our performance metrics and our benchmarks are documented here using real world testing conditions. Read more on why this is important here.
API Benchmarks
DeviceAtlas API has been clocked at speeds of up to 30 million detections/second on a high-end entry level server, using a randomized set of User-Agents.
The below benchmarks are based on tests with a single thread, and no caching, using a sample of 950,500 user agent strings extracted from our global network of thousands of partner websites. The C++ benchmarks were performed using a multithreaded test harness.
- A low spec 3GHz Intel Core 2 Duo with 4GB RAM running Ubuntu Linux for all APIs except .NET which was run under Windows 7.
- C++ case 1 - As above, dual core, multi-threaded
- C++ case 2 - High end 12 core 2.10 GHz Xeon E5-2620 v2 with 28Gb RAM) multi-threaded
API Type | Recognitions per second |
---|---|
PHP | 3,100 |
Python | 4,400 |
.NET | 26,700 (Windows 7) |
Java | 50,100 |
C++ | 1,700,000 (case 1, Low-end server) |
C++ | 30,000,000 (case 2, High-end entry level server) |