Drawbacks?

[jescalan]

Hey friends!

This library is super amazing and I am a huge fan of the work being done here. I was just curious if it might be helpful though to write out some of the drawbacks, if there are any, in the readme. For example, since it's using a lambda layer, I assume this means it will cut into the space available per lambda. The normal limit is ~250mb, and node image layers are ~100mb. Additionally, is there a performance penalty incurred when loading node through the layer on a cold boot? Is there anything else I'm missing?

Not trying to be a downer at all here, I just want to make sure that the tradeoffs are clear for anyone thinking about adopting 😁

[tjanczuk]

Thanks @jescalan! I will keep this issue open until we get around to updating the readme with those considerations. Quickly, however, there appear to be no performance penalty for cold start, but there is a performance penalty for creating the function itself - creating functions with layers takes more time, on the order of seconds.

[bestickley]

Have there been any official tests run to prove no perf penalty for cold start? I'm not worried so much about perf penalty creating the function, but would like to see numbers for cold starts. Thanks! Great project!