In the sign up process, it discourages using the primary srever. In the docs it talks about performance limitations. In contast Mastadon does not do this. Worse yet, most people think of a single centralized server. And this server only has some 500+ active users a month. Maybe 10 times that many lurkers?
While I love the Lemmy design, and the community looks great, I fear that it does not scale. Worse yet I suspect that the Activity Pub protocol is grossly inefficient for Reddit style applications. Maybe a publish subscribe message broker, like MQTT would be much more efficient.
So what is gong on here. Is this an issue? Am I correct to be scared about Lemmy scalability.
And just for context, I think you are totally on the right track, way ahead of all ofther reddit clones, but maybe just a little off course. Can you crank up performance?
Lemmy is written in Rust while Mastodon is written in Ruby, this likely explains the differences in RAM usage.
Maybe CPU/Disk will scale worse on lemmy because each page loads a lot of comments, while on Mastodon you rarely load more than a few toots at once.
I think you’re right, Mastodon’s CPU/RAM usage seems to be always the same. On Lemmy I can’t really try yet, as federation doesn’t seem to work for me. Once I can get content from other instances I’ll have more information.
I used to host a mastodon instance for just one user, me. It really struggled unfortunately in both speed and space requirements.
I wish they had an option to block all bots from the platform. I can block individuals and instances, but bots would still pop up like crazy.
From lemmy.ca:
USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMANDlemmy 17645 0.3 0.3 672368 98876 ? Ssl Aug03 154:51 node dist/js/server.jslemmy 20959 0.4 0.6 4127676 210048 ? Ssl Aug03 169:59 /home/lemmy/lemmy/target/release/lemmy_serverpictrs 32164 0.0 11.4 10599856 3769764 ? Ssl Jul28 18:50 /home/lemmy/pictrs/pict-rs/pict-rsiframely 20977 0.0 0.1 901704 55368 ? Ssl Jul08 3:25 iframelyMake of it what you will. :-)