New `Memory` backend, replace ETS?

[JuneKelly]

We seem to have a few issues with ETS backends, that are hard to make any better. Maybe a custom map-based storage server as the new default?

[teamon]

@ShaneKilkelly Can you share what are the issues with ETS backend?

I've noticed it's guarded by single GenServer process for both reads and writes - maybe allowing concurrent ETS reads from any process could improve the performance and make the ETS backend production-ready?

[robmadole]

@ShaneKilkelly just coming across this project. We are using ExRated right now on fontawesome.com and we are taking a look at switching to hammer.

And I second what @teamon is saying. I believe the performance issues are because of the synchronous nature of GenServer. ETS by itself is blisteringly fast but by wrapping it with atomic handle_call it slows everything down to essentially a single thread. We did this same thing on fontawesome.com for a bit until we read this article.

@ShaneKilkelly are you interested in a PR that takes the ETS backend and speeds things up for the critical path calls?

[JuneKelly]

Yup, I'd love to see that kind of PR! Thanks!

[robmadole]

@ShaneKilkelly take a look at #41 and see what you think

[epinault]

this is merged and solved