I’ve been considering launching a Lemmy server for a community lately, but am now back exploring piefed with the Lemm.ee shutdown.
I’ve read that Lemmy for some has been quite resource intensive and piefed was much easier on CPU and RAM. Is anyone able to share some metrics, say, for this server for reference?
It looks like documentation is better for piefed and the mod tools are better. I like the topics organization.
Downside is that the interstellar app didn’t seem to have that feature built in.
Also, my phone really does not like the word piefed lol.
But I’m starting to consider whether to trim a piefed instance instead of Lemmy. Is there anything I should know ahead of time? Such as unexpected gotchas or federation issues with Lemmy?
My Piefed testing instance only has two users so it’s hard to make a comparison with my Lemmy instance which has 180 monthly active users.
This is what my Lemmy instance looks like (180 monthly active users):
And my PieFed instance (2 users):
Again, not a fair comparison at all but I think they’re both pretty efficient with CPU/RAM. Postgres will eat as much RAM as you give it. I noticed that Lemmy will spike CPU every hour. On the other hand, PieFed spikes every time you reload the page but I don’t have any caching for PieFed enabled yet.
Anyway, I think another big concern in terms of resource usage is how big the DB grows. Technically, it grows infinitely. PieFed has this really nice feature that deletes old stuff over time per community (or that’s how I understand it at least) which is pretty nice:
PieFed has this really nice feature that deletes old stuff over time per community
On the one hand, that’s a cool feature and I can see how this could keep storage costs down.
On the other hand, access to old conversations is one of the main appeals of forums. If I knew my posts were guaranteed to disappear after X days, I’d reconsider posting there at all.
The instance admin can set it to different values on different communities. For example on piefed.social all the meme communities are set to either 6 months or 1 year. Other instances might have different settings.
I think it’s only the local copy. I could apply this setting to remote communities. It would be nice if Lemmy/Piefed/Mbin could fetch old remote posts on-demand instead of always keeping a copy of it.
thank you! that’s actually quite useful
The INSTALL.md says it’s a minimum of 2 CPU cores and 3GB of RAM, with 4 cores and 5GB+ RAM recommended. When that page was written, piefed.social used 100GB of disk space.
I run a single-user instance. It runs on a cheap 4core/8GB RAM vServer along a multitude of other services. My media storage on disk is just 8GB and I believe the postgres database has blown up to another 8GB. Obviously the numbers are going to increase once you or your users post more images and subscribe to more communities than I do.
I haven’t tracked CPU and RAM, but I think it’s pretty okay. Obviously there is some constant activity, since Fediverse servers exchange a lot of information and there are comments and votes coming in all the time.
thank you, that’s a good reference :)
When piefed.social first launched in early 2024 it had 2 CPU cores. That was fine until people subscribed to hundreds of communities (more than a single-user instance would typically have) so after a few months I increased it to 4 cores (8 GB of RAM). That was plenty until this week when the number of users doubled (now 500+ MAU) so now there are 8 cores. This seems like way more than necessary so when the excitement dies down a shift back to 4 might be possible.
You really really need a CDN and it needs to be set up right. See https://join.piefed.social/2024/02/20/how-much-difference-does-a-cdn-make-to-a-fediverse-instance/
Also tune PostgreSQL - https://join.piefed.social/2025/03/19/tuning-postgresql-for-piefed/
that is very good to know, especially the tuning. Thank you!
I just published some Cloudflare caching settings - https://join.piefed.social/2025/06/06/optimal-cloudflare-caching-config-for-piefed/