Performace tests investigation

[mantzas]

Is your feature request related to a problem? Please describe

There is a need to have performance tests for Patron. Let's investigate how to do this.

[tpaschalis]

I suppose one of the most basic ideas is to strategically place some benchmarks, and compare them between commits, right? There are some built-in tools like benchstat and benchcmp which we can use. We could also :

a) have a Makefile command that compares HEAD vs origin/master, and fails if the regression is over a threshold b) build a Github Actions workflow that does the same thing, and open-source it under the beatlabs organization c) use some off-the-shelf solution like gobenchdata, gobenchui or cob to keep track of performance over time

Some thoughts :

• It shouldn't be a hard requirement; complexity might increase due to adding features and not due to accidental performance regression
• These benchmarks should be relatively stable, so that don't end up comparing apples and oranges

[mantzas]

There is only the problem with infra. Everything runs on VM's so there is no guaratee on perfromance... tricky business

[tpaschalis]

Hmm, you're right. The benchstat tool actually reports differences only if they're statistically significant which should somewhat normalize this fluctuation. To do so it requires 4-5 runs per benchmark before and after to provide an answer which makes the benchmarks slower though. I'll keep looking into this

[mantzas]

I do not have a good answer on how to proceed with this. Maybe it is good enough since GitHub action has a specific instance type available for the CI. If they change the instance type to something else we will see an improvement/impact.