This is a benchmark to compare UUID generators, trying to replicate the experiment in this post by Johann Burkard.
Only UUID versions 1 and 4 are tested: version 3 and 5 are name-based; version 2 (DCE security) is not implemented in many libraries.
UUIDBenchmark is the benchmark for object generation, while UUIDToStringBenchmark is for string generation. Lower score is better (microseconds for each operation). For this test I used Eclipse Temurin 11.0.4, but tests with Oracle OpenJDK 17.0.2 are similar.
| Benchmark | Mode | Cnt | Score | Error | Units |
|---|---|---|---|---|---|
| UUIDBenchmark.testEAIOv1 | avgt | 16 | 0,013 | ±0,001 | us/op |
| UUIDBenchmark.testCommonsv4 | avgt | 16 | 0,042 | ±0,001 | us/op |
| UUIDBenchmark.testCommonsv1 | avgt | 16 | 0,052 | ±0,001 | us/op |
| UUIDBenchmark.testJUGv1 | avgt | 16 | 0,101 | ±0,001 | us/op |
| UUIDBenchmark.testJDKv4 | avgt | 16 | 0,363 | ±0,023 | us/op |
| UUIDBenchmark.testRiverv4 | avgt | 16 | 1,446 | ±0,044 | us/op |
| UUIDToStringBenchmark.testEAIOtoJava | avgt | 16 | 0,023 | ±0,001 | us/op |
| UUIDToStringBenchmark.testJUGv1 | avgt | 16 | 0,101 | ±0,001 | us/op |
| UUIDToStringBenchmark.testCommonsv1 | avgt | 16 | 0,109 | ±0,001 | us/op |
| UUIDToStringBenchmark.testCommonsv4 | avgt | 16 | 0,115 | ±0,002 | us/op |
| UUIDToStringBenchmark.testEAIOv1 | avgt | 16 | 0,122 | ±0,002 | us/op |
| UUIDToStringBenchmark.testJDKv4 | avgt | 16 | 0,370 | ±0,008 | us/op |
| UUIDToStringBenchmark.testRiverv4 | avgt | 16 | 1,604 | ±0,029 | us/op |