This stuff is never free anywhere but the scale of Big Tech insulates you from it somewhat.

I actually think it’s good to be reminded that our ability to communicate comes with a cost – in terms of resources, environmental impact, etc., not just money.

But, equally, it also makes the case that a system optimised to host hundreds of thousands of people on a single instance is not also somehow magically optimised to host just one person.

The latter is the problem I’m exploring with #SmallWeb.

That's interesting. I have my own instance with just a few other people on it, and I've been trying to get an understanding of what kind of resources does running Mastodon require.

It seems that it is quite a resource hog. I wonder if something can be done to further optimize the federation.

it feels like a lot of what it must be doing is the same as all the other instances out there are doing, give or take different overlapping sets of data.

That doesn’t feel very efficient, but it also feels like a space where a cooperative caching layer could sit; one that is contributed to by all, without needing to being owned by anyone. Sort of like a CDN in that it’s used if available, but not required for the server to work.

I think that this sort of problem scales out much better if you are on a shared instance.

The way I understand it, instances share content with each other when needed. So if you are following 4k people, but 3k of those have already been followed by others on your instance, you share that traffic :>

Plus, instance-internal follows are cheap, and you get a more even usage of your resources rather than "mostly nothing until someone very popular posts"

I think this is the biggest potential problem (very real problem we’re starting to see) with the fediverse in general… a large number of connections grows the federation and communication required exponentially, not linearly. This is difficult enough to deal with in a centralized network, but in a distributed one it becomes an even more complex challenge.

Scaling this thing is going to be fun.

oh boy 😲
that 19 GB is too much, my VPS has only 24 and I'm hosting everything I can on it.

there are only 2 people on my instance currently and I'm mostly following people I knew from Twitter who have migrated recently.

is it the amount of input/output network traffic that raises the costs so much?

On my instance I delete about 1GB a day in unnecessary media attachments that are exchanged, I guess that number is much higher for you?

Not knowing anything about the back end of a mastodon server and how it scales, what's the impact on server load when you go from, say, 20k followers to 200k to 2M?

As the fediverse scales up, are we looking at a world where it's fair to ask the users with the biggest followers to shoulder more of the hosting costs?

From my understanding is it would increase your database usage mostly. Bandwidth would spike too I suppose depending on how much talk you get from the servers.

What type of hosting are you using?

As an instance operator, I decided to offload objects to S3/Cloudfront and Email to SES (mostly to avoid dealing with emails getting flagged as spam).

this is what I was curious about and figured might be an issue. Being concerned about this has already earned me some scalding in the network.

On a personal level I'm wondering what I could feasibly do at home and possibly tunnel through Cloudflare.

On a large scale, I see issues

Thanks a bunch for posting this. I was wondering how the single-user instance scaling would work out.

I've been wondering about the single-user vs multi-user instance question from a performance standpoint ever since I read your arguments for smaller instances.

I seems I will have to read the ActivityPub spec at some point.