Microbenchmark used to compare /dev/random and /dev/urandom for UUIDs
The benchmark creates 1 million UUIDs in parallel with either /dev/random or /dev/urandom.
If the blocking from /dev/random is really altering performance, we should see a notable difference.
Results on my ryzen 5 3500U 5.16.14-1-MANJARO :
| Benchmark | Mode | Cnt | Score | Error | Units |
|---|---|---|---|---|---|
| MultithreadedUuidsBench.withRandom | avgt | 5 | 4816.977 | ± 64.741 | ms/op |
| MultithreadedUuidsBench.withUrandom (8 threads) | avgt | 5 | 4833.015 | ± 49.046 | ms/op |
My interpretation : There is no difference with few threads.
Maybe with 64 threads there are, but I doubt.