What exactly is the cost of self hosting a Lemmy instance? Understandably you would want a powerful server, but that would be just a single one time purchase.
Where does the rest of the cost come in? Does it require more than a 2 gigabit connection and thus require a data centers 10 gigabit connection?
If I could run an instance on 2 gigabits and spending a 1-3 thousand on a server then I’d be interested in giving it a try.
I’m hosting a personal instance in my home rack on a VM of 1 CPU / 2GB RAM. My only concern is to see how fast the disk usage grows over time (years?) no idea yet if lemmy cleans up old content or not. I’d imagine not since that defeats the purpose.
Depends how many people you are hosting for. But what you describe seems overkill, I think Lemmy.ml just upgraded to a 6 core cpu and 32gb of ram and they have 30k+ members.
That’s a pretty good benchmark. Thank you!
Depending upon how many users you are looking to host, Lemmy instances are not very taxing. Most instances are on quite modest hardware.
I host an instance for myself. I have subscribed to many communities (10-20), and I run it on a 1 CPU + 1 GB RAM DigitalOcean droplet. However, the Lemmy instance was OOM-killed already once, and I expect that I have to upgrade eventually.
The droplet costs $6/month.
would swap help?
“If you need to use the swap, you’re doing it wrong” – That’s what I learned long ago. And it has held up so far.
Interesting. I’ve never heard that. I use swap all the time and it’s saved me from OOM scenarios. I’m currently limited on RAM so maybe it makes more sense for my situation.
Right? Like, my “first” machine had 512MB of ram in an era where most people were running 4GB. SWAP made more modern distributions possible for me. I mean, then again, that wasn’t so much a choice than the harsh reality of growing up broke!
(My ACTUAL first machine was more like 64MB, but I never did much with it)
zramand other compressed swap approaches can help too (with less of a performance hit) although I use real swap as a fallback. Some would recommend usingzswapin that case, but I still want compression in ram to be heavily prioritised but YMMV.
Not much at all.
For a basic instance running <1k registered users, it seems the lemmy.ml instance survived a while with a quad core and some middling ram on a VPS. Really, connection bandwidth and/or CPU is not the bottleneck with Lemmy since most heavy image/video content continues to be linked externally, it’s the database management to keep all comments and posts sorted correctly that takes resources. So you need storage IO and memory more than anything.
Sure if you want to grow as large as lemmy.ml you will probably be needing a “dedicated server” with high expenses. However, I’d bet any old generic office PC, or even a Pi, can run most small instances easily if you have access to a 2-gig symmetrical connection.
