This repo contains the full code of my blog entry.
- Run
npm install. (This is a NodeJS app). - Create a file named
.envin the root folder.
The .env file content:
STARTING_LENGTH=10 # Starting number of concatenations.
MULTIPLIER=10 # Multiply <STARTING_LENGTH> by this number in every iteration.
ITERATIONS=5 # Number of iterations to be tested with.
#DISABLE_BUFF_TO_STRING=true # Disable converting `Buffer` to `String` in `service_v3.js`.To be fair, the app node <file.js> must be ran using the flag --expose-gc,
so we can use global.gc() to do a full garbage collector run before test, so the garbage collector does not run
in the middle of the tests.
There are two 'different' ways of running it:
node --expose-gc index.jsnode --jitless --expose-gc index.js
The way #2 actually disable the jit compiler with the flag --jitless.
Environment: DISABLE_BUFF_TO_STRING --> If you set it to true (you can just uncomment the line by removing the leading #)
you can see that the services actually runs faster, but the result is not the same, but can be useful depending on your scenario.
- CPU: I7-9700K
- RAM: 32GB - 3200MHz
- Windows 10 (20H2)
- Node version: v16.5.0
STARTING_LENGTH: 10 | MULTIPLIER: 10 | ITERATIONS: 5 | DISABLE_BUFF_TO_STRING: true
| service version | iteration #1 | iteration #2 | iteration #3 | iteration #4 | iteration #5 |
|---|---|---|---|---|---|
| V1 - Concatenation | 0ms (27.14 MB) | 0ms (27.73 MB) | 0ms (27.64 MB) | 0ms (29.57 MB) | 20ms (40.26 MB) |
| V2 - File system as buffer | 26ms (27.44 MB) | 10ms (28.41 MB) | 80ms (36.20 MB) | 735ms (97.33 MB) | 6745ms (708.92 MB) |
| V3 - Direct Buffers | 0ms (27.45) | 0ms (27.92 MB) | 1ms (32.91 MB) | 7ms (63.60 MB) | 74ms (374.65 MB) |
| V4 - Using Arrays | 0ms (27.72 MB) | 0ms (27.91 MB) | 2ms (32.91 MB) | 12ms (63.67 MB) | 127ms (376.08 MB) |
Those results are extracted from logs/standard_run.txt.
Memory were measured with process.memoryUsage.rss() / 1024 / 1024.
If you need more test results, please, test it by yourself 😊.