refactor: better python performance

[Lewiscowles1986]

So I tried a few things, which get the minimum time in k6 down to microseconds for a request. They set slightly more aggressive timeouts, use pydantic to marshal the response from the POST endpoint to the DB. and set a latency on every request DB calls of 1 second. (the set local DB call, actually creates some overhead)

Ive not re-run the entire test, instead I've contributed a k6 test, which I used to verify on my local machine; within the also modified docker-compose yaml.

The migrations are still manual; and you'll probably notice I tried increasing server connections. I Think adjusting the min_size of db connections could be another good way to gain performance.

I'm pretty sure it still wont top bun, but you could get a few more thousand requests per second

[Lewiscowles1986]

I did not run as a pod, instead I just ran against docker

https://github.com/antonputra/tutorials/pull/500/files#diff-67c6f0510b3d66f24c698033facfaf94ca0e45b597c3db21c56d7a35dfda3ed2R16

[emerson-proenca]

Do you believe running Pypy would improve Python performance?

Pypy + FastAPI = fastapi/fastapi#3944
Pypy speed = https://speed.pypy.org/

[Lewiscowles1986]

Maybe @emerson-proenca; If you make a PR I'd love to run it. I've not had a lot of experience with it, so I consider it to be esoteric at the moment; but maybe it's very easy and I'm just unclear about approach.

I do know AI was telling me I was barking up the wrong tree with this, but then when I looked at all my latencies, they seemed to support me, more than AI review of my work.

[Lewiscowles1986]

Here are some logs from K6 test included with PR, run on an M2 Air laptop

          /\      |‾‾| /‾‾/   /‾‾/   
     /\  /  \     |  |/  /   /  /    
    /  \/    \    |     (   /   ‾‾\  
   /          \   |  |\  \ |  (‾)  | 
  / __________ \  |__| \__\ \_____/ .io

     execution: local
        script: ../script.js
        output: -

     scenarios: (100.00%) 1 scenario, 100 max VUs, 1m0s max duration (incl. graceful stop):
              * default: 100 looping VUs for 30s (gracefulStop: 30s)


     ✓ is status 201

     checks.........................: 100.00% ✓ 137626      ✗ 0     
     data_received..................: 44 MB   1.5 MB/s
     data_sent......................: 26 MB   878 kB/s
     http_req_blocked...............: avg=2.77µs  min=0s     med=1µs     max=18.46ms p(90)=2µs     p(95)=2µs    
     http_req_connecting............: avg=1.22µs  min=0s     med=0s      max=6.77ms  p(90)=0s      p(95)=0s     
     http_req_duration..............: avg=21.58ms min=6.09ms med=18.25ms max=1.45s   p(90)=28.56ms p(95)=37.09ms
       { expected_response:true }...: avg=21.58ms min=6.09ms med=18.25ms max=1.45s   p(90)=28.56ms p(95)=37.09ms
     http_req_failed................: 0.00%   ✓ 0           ✗ 137626
     http_req_receiving.............: avg=16.37µs min=4µs    med=10µs    max=5.15ms  p(90)=30µs    p(95)=40µs   
     http_req_sending...............: avg=5.25µs  min=1µs    med=3µs     max=4.96ms  p(90)=8µs     p(95)=12µs   
     http_req_tls_handshaking.......: avg=0s      min=0s     med=0s      max=0s      p(90)=0s      p(95)=0s     
     http_req_waiting...............: avg=21.56ms min=6.02ms med=18.23ms max=1.45s   p(90)=28.53ms p(95)=37.08ms
     http_reqs......................: 137626  4585.523346/s
     iteration_duration.............: avg=21.79ms min=6.47ms med=18.45ms max=1.45s   p(90)=28.77ms p(95)=37.32ms
     iterations.....................: 137626  4585.523346/s
     vus............................: 100     min=100       max=100 
     vus_max........................: 100     min=100       max=100 

[Lewiscowles1986]

Apologies, I realised after re-running the base-line I'd committed something I should not have. I've commented it out, but basically, I doubled the number of queries being run as part of a failed experiment and then accidentally git add . it. The rest is done now

[antonputra]

Thank you, @Lewiscowles1986, for the PR. Sorry I don't have clear guidelines for pull requests. Would you be able to just create a fastapi-app-v2 folder and keep the original as is? It's just easier to compare in the future.

[Lewiscowles1986]

To be honest, it merging isnt' the most important thing as you'll only get at most a few thousand requests (not per second, overall 😄 )

Thank you for the tutorial; if you'd still like the changes I'll make them

[antonputra]

@Lewiscowles1986 Yes please, I'll keep it. When it's time to refresh the benchmark, I'll go over and test everything again.