this post was submitted on 11 Apr 2024
59 points (98.4% liked)
/0
1561 readers
92 users here now
Meta community. Discuss about this lemmy instance or lemmy in general.
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
It's more to do with the lemmy itself as a platform handles than bandwith. Basically, there is only one channel between any two instances, and it's serial, and each step requires multiple handshakes to complete. Add in geographic distance making those handshakes take a significant part of a second to complete, and you end up with a single channel that gets flooded.
Blahaj.zone is 1.3 million activities behind on lemmy.world for example, but we're not behind on any other instance, because those channels don't hit capacity. Now, if we could use multiple channels at once to talk to lemmy.world, we wouldn't have a problem, but lemmy isn't built for that at the moment
Are you in the matrix general chat?
I'm in the lemmy.world admin back channel, but that's it as far as lemmy.world goes. @ada:chat.blahaj.zone
This one is the most active: https://matrix.to/#/#lemmy-support-general:discuss.online
I left that channel when one of the lemmy devs decided to take a swipe at our instance for the way we handled transphobia on another instance
Oh :(
Lemmy.world is really too big. We just moved !casualconversation@lemmy.world to !casualconversation@lemm.ee to try help with those issues.
Hopefully other communities will do the same, but the issue doesn't seem to be very well known.
By the way, thank you for your post about centralization of communities a while back, I used it in the post to explain to the community why we were moving: https://lemmy.blahaj.zone/post/10810804?scrollToComments=true
The idea behind federation was that you don't get single instances so large that the single sync channel gets flooded. You're meant to have more smaller instances spread out that then federate with one another like a mesh- only one copy of data needs to be sent to each instance to serve additional thousands of users without further load.
That said, realistically the fediverse needs single instances with 10-30k users to really thrive, and L.W hasn't even hit 10k actives... so there does need to be some batch improvements in the backend still.
Having the ability to "merge content" from multiple instance communities would help a lot too. People congregate on L.W both from inside and outside the instance because it's big and has the most content. But if you have multiple communities (e.g, search "memes" and see how many different memes communities there are) the others with less content get ignored. Being able to merge those from a user viewing perspective, so you would just have a "memes" group that sees federated content from all the memes communities, could reduce the need for L.W to be all encompassing.
Couldnt they batch them up? Im not a technical person but this seems solvable.
Yep. Batching and/or multiple parallel channels per remote instance would solve it.