lml

My kbin instance’s data (text data, that is) probably takes up less than 8 GB right now, and I’ve had it running about two weeks. Media storage (which I do through S3) is around 5 GB so far. Kbin does do media mirroring different than Lemmy though (I think), so YMMV. I think Lemmy mostly links to the original instance’s media object as the source.

The main thing I found eating storage on my server was a lot of old Docker images (and volumes) from me trying to get everything up and running. If you are using Docker you could try doing a docker system prune --all to get rid of unused images/build caches (anything that isn’t running currently).

I think that would be worth it, yeah. Of course if you are hosting it on your home network there will be some added security concerns (and that might make it better to only allow signups to friends/friends of friends/etc). The way I see it is that some instances are going to host the largest communities, and therefore those instances are going to need to handle all of the incoming/outgoing updates to posts in those communities. Right now they can’t do that reliably and push updates out to all of their users’ devices.

So in the long run I think having small/medium instances (say a couple hundred, not tens of thousands of users) will be the way to grow. These smaller communities can push updates to their smaller user count reliably, and then have more resources to handle federated content coming in and going out. I think scaling for the incoming/outgoing federation requests would be easier than for direct user activity. Federation stuff can be queued and then spread over time, but user requests cannot be.

Good point. Nothing against the larger instance owners of course. If my little instance got super popular somehow (like being recommended in guides on how to join lemmy/kbin), and thousands of users got in per day, I could see issues happening just like this. I don’t know the ins and outs of tuning this software for performance at scale, and I know I couldn’t learn it fast enough if my instance faced very fast growth like lemmy.world has.

I think admins are going to need to turn registrations off periodically, as they scale their hardware (and their knowledge) to run it for more users effectively.

This is the problem we are having with fast growth on a few select communities. The largest servers are being bogged down simply because the software has not been tuned for these large types of instances yet. ActivityPub works best (in it’s current state) by spreading users over smaller/medium sized instances. Folks need to take a look at other instances (and I agree it is hard to find them for a newcomer). You can look at https://fedidb.org/ to look at instances that have been indexed running kbin, lemmy, and other software.

Joining a smaller instance means that your server is not being bogged down by tens of thousands of other users trying to pull updates at the same time. You can still see the content from other instances, and in many cases it is more reliable because your smaller instance actually has the resources to handle pulling in the posts you want to see. In the future I am sure instances like lemmy.world will be able to handle the traffic smoothly, but for now the best way to ensure stability is to join a smaller instance.

(Plug for my instance: https://remy.city, a general purpose Kbin instance. I set it up for personal use but anyone is free to join me in using it. I have defederated from the more alt-right communities like lemmygrad and exploding-heads, and from lemmynsfw.com because of content hosting concerns. I’m open to suggestions on others.)