Benchmark of popular async queue libraries (and also mine) in workerd.
All libraries run the exact same test, so promise-queue and p-limit are not penalized by having to use Promise.all (they don't provide a promise that resolves when the queue is empty).
| Library | Requests/sec | Total (sec) | Average | Slowest |
|---|---|---|---|---|
| @henrygd/queue | 816.1074 | 1.2253 | 0.0602 | 0.0864 |
| promise-queue | 647.2809 | 1.5449 | 0.0759 | 0.1149 |
| fastq | 336.7031 | 3.0877 | 0.1459 | 0.2080 |
| async.queue | 198.9986 | 5.0252 | 0.2468 | 0.3544 |
| queue | 85.6483 | 11.6757 | 0.5732 | 0.7629 |
| p-limit | 77.7434 | 12.8628 | 0.6316 | 0.9585 |
pnpm i && pnpm devSends 100 requests to warm up the server, then 1000 requests to benchmark using oha.
Changing the warm up path doesn't make a difference. It compiles to a single file and each test is standalone. I tested with fastq and got a longer time, so warm up is left on henrygd-queue to have a more uniform warm up time.
# @henrygd/queue
oha --no-tui -n 100 http://localhost:8787/henrygd-queue > /dev/null && oha -n 1000 http://localhost:8787/henrygd-queue
# promise-queue
oha --no-tui -n 100 http://localhost:8787/henrygd-queue > /dev/null && oha -n 1000 http://localhost:8787/promise-queue
# async.queue
oha --no-tui -n 100 http://localhost:8787/henrygd-queue > /dev/null && oha -n 1000 http://localhost:8787/async-queue
# fastq
oha --no-tui -n 100 http://localhost:8787/henrygd-queue > /dev/null && oha -n 1000 http://localhost:8787/fastq
# queue
oha --no-tui -n 100 http://localhost:8787/henrygd-queue > /dev/null && oha -n 1000 http://localhost:8787/queue
# p-limit
oha --no-tui -n 100 http://localhost:8787/henrygd-queue > /dev/null && oha -n 1000 http://localhost:8787/p-limit