@lain Hey bruh, i was curious a bit, do you have some stats about throughput of the biggest pleroma instances you know? Like, how many messages/hour handled, on which hardware ?
Timeline
Post
Remote status
Replies
6
@Arwalk sorry, no idea, but i'm interested too. might make sense to set up a few simulations.
@lain Damn, we're discussing (dreaming) on doing some server software that's notably hard to handle but i'm pretty sure elixir would be quite ideal for that, so i wanted to get a rough idea of that. We're not planning on doing social network software but the level of activity would be relatively close imho.
Thanks anyway, if you get any info on that i'd be interested!
Thanks anyway, if you get any info on that i'd be interested!
@Arwalk if you do any kind networking stuff i'd either do go or elixir. elixir is much easier to work with and reason about and can get you a million connections. https://www.phoenixframework.org/blog/the-road-to-2-million-websocket-connections
@lain @Arwalk a 7 day example of subscribing to every single relay, pleroma hardly uses any cpu/memory at barely any load and ~1gb, but postgresql really really struggles if you aren't running it on a raid of plenty of ssd/ram, because if anything stalls it errors out hard
@lain @Arwalk it also helps that everything networking is ridiculously well fleshed out and trivial, with everything else possible to fill out with the help of https://dashbit.co/blog/why-elixir-best-language-for-ai because everyone's code is written mostly the exact same way