Benchmarks
This page provides guidance figures for detection speeds, memory usage and startup times for Pattern and Trie algorithms. Pattern is more memory efficient and can be run directly from disk while Trie requires considerably more main memory but delivers millions of detections per second. For more information check out the how device detection works page.
Results stated are detections per core on a quad core i7 2.2GHz CPU.
Pattern Benchmarks
Below is a table of performance metrics relating to the Python Pattern API. It shows the detection speed for a single request for each data set and where applicable also the mode of operation used.
Lite | Premium | Enterprise | |
---|---|---|---|
Detections Per Second | 55865 | 49751 | 43668 |
Time Per Detection (ms) | 0.0179 | 0.0201 |
0.0229
|
Lite | Premium | Enterprise | |
---|---|---|---|
Startup Time (ms) |
16.63
|
29.95
|
34.35
|
Average Memory Usage (Mb) |
54
|
109
|
147
|
Hash Trie Benchmarks
Below is a table of performance metrics relating to the Python Hash Trie API. It shows the detection speed for a single request for each data set and where applicable also the mode of operation used.
Single Thread | Lite | Enterprise |
---|---|---|
Detections Per Second | 559,394 | 572,671 |
Time Per Detection (ms) | 0.001788 | 0.001747 |
Lite | Enterprise | |
---|---|---|
Startup Time (ms) | 63 | 79 |
Average Memory Usage (Mb) | 116 | 142 |