autocannon-0003.out

# Hyper-Express

$ autocannon http://localhost:3001/api/check --renderStatusCodes --debug --warmup [ -c 2500 -d 30 ] -c 2500 -d 30 -p 4

┌─────────┬────────┬────────┬────────┬────────┬───────────┬──────────┬────────┐
│ Stat    │ 2.5%   │ 50%    │ 97.5%  │ 99%    │ Avg       │ Stdev    │ Max    │
├─────────┼────────┼────────┼────────┼────────┼───────────┼──────────┼────────┤
│ Latency │ 267 ms │ 447 ms │ 622 ms │ 645 ms │ 427.02 ms │ 76.18 ms │ 708 ms │
└─────────┴────────┴────────┴────────┴────────┴───────────┴──────────┴────────┘
┌───────────┬─────────┬─────────┬─────────┬────────┬──────────┬──────────┬─────────┐
│ Stat      │ 1%      │ 2.5%    │ 50%     │ 97.5%  │ Avg      │ Stdev    │ Min     │
├───────────┼─────────┼─────────┼─────────┼────────┼──────────┼──────────┼─────────┤
│ Req/Sec   │ 21023   │ 21023   │ 24207   │ 202879 │ 48620.71 │ 49885.69 │ 21016   │
├───────────┼─────────┼─────────┼─────────┼────────┼──────────┼──────────┼─────────┤
│ Bytes/Sec │ 14.3 MB │ 14.3 MB │ 16.4 MB │ 138 MB │ 33 MB    │ 33.8 MB  │ 14.2 MB │
└───────────┴─────────┴─────────┴─────────┴────────┴──────────┴──────────┴─────────┘
┌──────┬────────┐
│ Code │ Count  │
├──────┼────────┤
│ 200  │ 826472 │
└──────┴────────┘

Req/Bytes counts sampled once per second.
# of samples: 17

836k requests in 35.7s, 560 MB read

autocannon-0004.out

# ExpressJS

$ autocannon http://localhost:3002/api/check --renderStatusCodes --debug --warmup [ -c 2500 -d 30 ] -c 2500 -d 30 -p 4

┌─────────┬────────┬─────────┬─────────┬──────────┬────────────┬────────────┬──────────┐
│ Stat    │ 2.5%   │ 50%     │ 97.5%   │ 99%      │ Avg        │ Stdev      │ Max      │
├─────────┼────────┼─────────┼─────────┼──────────┼────────────┼────────────┼──────────┤
│ Latency │ 643 ms │ 2597 ms │ 9645 ms │ 11307 ms │ 3328.36 ms │ 2016.34 ms │ 11482 ms │
└─────────┴────────┴─────────┴─────────┴──────────┴────────────┴────────────┴──────────┘
┌───────────┬─────┬──────┬─────────┬─────────┬─────────┬────────┬────────┐
│ Stat      │ 1%  │ 2.5% │ 50%     │ 97.5%   │ Avg     │ Stdev  │ Min    │
├───────────┼─────┼──────┼─────────┼─────────┼─────────┼────────┼────────┤
│ Req/Sec   │ 0   │ 0    │ 3557    │ 4819    │ 3580.14 │ 984.83 │ 1024   │
├───────────┼─────┼──────┼─────────┼─────────┼─────────┼────────┼────────┤
│ Bytes/Sec │ 0 B │ 0 B  │ 2.86 MB │ 3.87 MB │ 2.87 MB │ 791 kB │ 822 kB │
└───────────┴─────┴──────┴─────────┴─────────┴─────────┴────────┴────────┘
┌──────┬────────┐
│ Code │ Count  │
├──────┼────────┤
│ 200  │ 107378 │
└──────┴────────┘

Req/Bytes counts sampled once per second.
# of samples: 30

121k requests in 30.58s, 86.2 MB read
1k errors (496 timeouts)

fetch-get.js.result

$ env FETCH_TOKEN="MY_TOKEN" node ./src/fetch-get.js
/
hyper-express x 1,446 ops/sec ±4.02% (73 runs sampled)
express x 462 ops/sec ±3.36% (69 runs sampled)
Fastest: hyper-express

/api/check
hyper-express x 1,740 ops/sec ±2.32% (75 runs sampled)
express x 927 ops/sec ±1.79% (77 runs sampled)
Fastest: hyper-express

/api/uploads/0
hyper-express x 599 ops/sec ±2.66% (72 runs sampled)
express x 457 ops/sec ±3.32% (73 runs sampled)
Fastest: hyper-express

/api/albums/0
hyper-express x 369 ops/sec ±2.72% (72 runs sampled)
express x 335 ops/sec ±3.12% (72 runs sampled)
Fastest: hyper-express

/api/album/get/GUb5qeQ1
hyper-express x 478 ops/sec ±3.41% (72 runs sampled)
express x 343 ops/sec ±3.59% (70 runs sampled)
Fastest: hyper-express

fetch-get.js.result is result of src/fetch-get.js in https://github.com/BobbyWibowo/perf-playground
Autocannon results are likely more reliable, since the tests performed with the aforementioned script will do them one-by-one in sequence (i.e. exactly 1 concurrent connection)
Nonetheless, even in that context we can see considerable performance improvement, so hyper-express simply wins all around

/api/check route is only a simple JSON response of some config options, so it's more of each web servers' actual performance, I'd think

/ route is also considerably faster because hyper-express branch comes with persistent internal cache for Nunjucks front-end pages

The other /api/* routes involve some DB queries, which are identical on both branches, so improvements came from the web servers alone