In which I write about “Basic Infrastructure”: https://www.tbray.org/ongoing/When/202x/2024/08/13/Basic-Infrastructure
Opening para: “Recently, I was looking at the infrastructure bills for our CoSocial co-op member-owned Mastodon instance, mostly Digital Ocean and a bit of AWS. They seemed too high for what we’re getting. Which makes me think about the kind of infrastructure that a decentralized social network needs, and how to get it.”
@timbray OVH would do the trick and would definitely give you that hour of downtime on the regular.
The issue is more how to deploy+configure a database and other necessary dependencies on a random Linux host (if you don’t want to use your host provided db instance). Not sure how to make that truly easy and turnkey.
@timbray I host GoToSocial on a DigitalOcean VM. I have 2 users. It’s a minimal resource Docker container, with data in a SQLite DB. I could put the assets in s3-compatible storage, but the VM has plenty of space to spare so why bother?
I’d love to see more work toward easy self-hosting small instances for micro-communities that can federate to the larger community.
@timbray I agree with Wes in your comments that Mastodon is inefficient, but I would rather see competitors than a rework - it’s needed to keep the fediverse the fediverse and not Mastodon
If we are looking at competitors, it would also be interesting to look at different protocols. Ejabberd can serve millions of XMPP sessions on a cheap VPS. We don't need to squint our eyes that hard to treat ActivityPub as a messaging server, it should be able to reach close to it.
Also, delegate a lot of stuff to the edge: https://raphael.lullis.net/a-plan-for-social-media-less-fedi-more-webby/