mekka okereke :verified:
"In 2024, it cost $N to run a Mastodon instance with ~1000 active users for a year. By the end of 2025, that figure had dropped to $0.5X, due to the launch of X, Y, Z."
What are X, Y, and Z?
In terms of cost, consider systems administration and moderation as paid work, even if the admin is not paying themselves for either of these things.
Tim Bray
@mekkaokereke
X: Cheap competitors in generic SaaS offerings like blob storage and CDN.
Y: IFTAS, maybe.
Z: (stretching here) AI (but not genAI) models to help moderators with traffic the way it helps radiologists with images.
craignicol
@mekkaokereke I suspect there's more than one startup who will offer AI moderation as one of the three - there must be enough public social media data to train a decent spam filter.
Mastodon/ActivityPub extensions for common platforms so anyone with an existing community can extend it into the Fedi without new hardware
A Mastodon fork optimised for shared hosting so a small admin team can run hundreds of instances on a handful of servers
Louis Ingenthron
@mekkaokereke The obvious answer would be centralization of services. With groups like MastoHost using shared resources for lots of servers, the economies of scale drops the overall per-server costs. If there's enough demand, that would even create competition to lower the price further.
Emelia 👸🏻
@mekkaokereke current estimates of yearly cost per account is $0.30 to $0.80 based on infrastructure, storage, etc. From what I've seen.
I'm pretty sure @esk or @dma worked out the numbers for Hachyderm too.
Tom Walker
@mekkaokereke The server is quite "heavy" so it could probably be something like
X - code speed improvements
Y - reducing cache sizes or something
Z - switching from Rails to anything that isn't dog slow
Perhaps not what you had in mind but for a cheap-ish VPS it really just needs to be lighter weight and faster
Samir Al-Battran
@mekkaokereke
X = Fediverse discovery project
Y = Mastodon implementation of it
Z = Getting rid of relays and local copy of remote data
Fediverse discovery project will actually be a huge advantage.
I don't know if it will be complete in 2025, but if it did then it's a game changer (Also assuming Mastodon's architecture takes advantage of it)
The existing design is not efficient, and also you end up missing most of the conversations (esp on smaller instances with 1K users)
Ramon Fincken 🇺🇦
@mekkaokereke redis and CDN?
Ramon Fincken 🇺🇦
@mekkaokereke and firewalling anti spam
Renaud Chaput
@samir @mekkaokereke we are doing this under a grant from @ngisearch and the agreed plan is to have it done by next June. The spec work is mostly done (unfortunately we did not get much feedback) and we hope to have a first implementation for trends (our first capability) in 2 months, both the « provider » side and the Mastodon implementation
Renaud Chaput
@mekkaokereke if we focus on cost, then shared moderation and shared storage. Those 2 things build on our FASP idea that we are currently actively working on.
I see a lot of people pointing technical things like switching from Rails or some other brick, but those are really not the issues. At least from my experience running instances with many hundred thousand users.
Brandon Jones
@mekkaokereke I've heard ActivityPub is pretty chatty as far as protocols go. Wonder how much headroom there is for reducing server costs purely through protocol improvements?
(That does nothing to reduce the human costs for moderation, but I think we all know tech folks are more likely to go for the easily quantifiable tech solutions first.)
gkrnours
@mekkaokereke I wonder if shared moderation team could help. Like a moderation panel that could handle report for multiple instances and a few small instances that help moderate each other instance using such a tool. This way, a dozen instance have one mod available 2h a day, instead of being unmoderated 22h a day, could be moderated all around the clock.
In the past, spam filter have been used to classify text content. Maybe it could be done for triage in moderation.
Jenniferplusplus
@mekkaokereke X is an activity transfer relay. Similar to mail transfer relays, in SMTP.
Y is a shared media storage pool.
Z is automated spam filtering.
Eashwar
@mekkaokereke I feel like pooling storage could be very useful since a lot of servers are storing copies of the same media (see jortage.com). Media storage, CDN, and transfer bandwidth make up a huge amount of cost in running servers.
mhoye
@mekkaokereke Pubsub-style shared moderation, shared-cost media caching, and normalizing media-content expiry.
Ian Smith
X = content addressable distributed file shares (i.e., ipfs, etc) for storage to diffuse the cost of cloud storage across all members on the server.
Y = content distribution cooperative to increase buying power worldwide for data connectivity.
Z = p2p client update propagation between members of the same server to reduce the hit rate on the servers themselves.
The Animal and the Machine
@mekkaokereke
With the greatest respect, I don’t think cost is the issue, I think it’s revenue.
The cost per user should be made public and those who both can and want to, can donate for it. No perks.
Alternatively, we could use the free market to create many special interests, from black communities to zoology. These host people for a standard fee and a code of conduct. The low fee allows for solid moderation and an account is used as an algo. Follow for boosts. We could do this today.
These are additive to the existing free ones, not replacing. If they scale to be only $1 a month, that is an affordable option for most.
I think this will happen when we get the right nerdy community leader ready to spend the effort to do it.
mekka okereke :verified:
Agree that "finding other ways to pay for things," is important.
But I do think that making things cheaper is important too, so that there's less stuff to have to pay for.
:PUA: Shlee fucked around and
@mekkaokereke reading these posts.. I hope 2025 is the year of shared services on the fediverse.
shared community / co-operation services are not centralisation you zealots. Don't @ me.
Rich Felker
@craignicol @mekkaokereke No one is going to use an instance that abdicates moderation in favor of AI shit. Pattern recognition that cannot see and understand outside context cannot accurately moderate.
Rich Felker
@mekkaokereke X, Y, and Z are all downsampling giant jpegs and pngs to <100k jpegs.
Rich Felker
@mekkaokereke For both local and remote posts, only show severely reduced size images direct in-feed. On click to expand, show full quality if local (possibly with upload size limits) and either iframe the remote from originating instance if permissions allow it, or show moderate size cached copy.
Instances should start setting policy headers to allow such iframing for known friendly federated peers.
Rich Felker
@mekkaokereke Or even just remote src img rather than iframe, but I'm not sure if you can access control that with referer possibly being suppressed (as it should be).
Rich Felker
@craignicol @mekkaokereke The horror of AI shit doing moderation is not false negatives but false positives. No one wants to be where they'll get randomly banned by opaque processes.
craignicol
@dalias @mekkaokereke people accept instances with barely any moderation (well, a lot of instances defederate them, but they remain popular) and I'm sure threads will stay. It's not something I'll be comfortable with, but it's a trade-off people who don't feel threatened here might accept.
Rich Felker
@craignicol @mekkaokereke On Facebook they stay because it's holding their social relationships hostage. On fedi there's no reason to stick with a shit instance that's at high risk of taking your social relationships away when there are plenty of good ones.
craignicol
@dalias @mekkaokereke I know people who had temporary bans from Facebook for unknown reasons but got reinstated and are still there. I don't see why that behaviour wouldn't be replicated here. If moderation isn't something you think about much, why would it matter to you if it was a faceless AI or a faceless minimum wage worker (or volunteer) that's deleted that one message in support of Harris?
Andre
@mekkaokereke Just about every non-obvious improvement to Mastodon or ActivityPub would seem to carry an additional trust requirement, either implicit or explicit.
One of the great joys of the current protocol is anyone with a little motivation can spin up an instance. The only gates are a DNS entry and some sort of compute and an internet connection.
Much more and things start to require trust relationships between servers which crushes out the little guy.
Raccoon🏳️🌈
@lispi314 @craignicol @dalias @mekkaokereke
I want to note something, as the person in charge of moderation on a major server: there is a lot that can be done and hasn't with non-AI based automated moderation. Why is there not a feature built into Mastodon to...
- automatically report posts with racial slurs when they show up?
- automatically report when a user posts the same link several times in a row?
- automatically report when a user from your instance has been mentioned or linked on a hashtag like FediBlock?
- remind users of the report button when they get PMs that might be potential harassment?
We don't need AI. It would be wasteful, very costly, and likely have serious trust and reliability issues. We just need better systems to be built in by default to help us volunteers catch things quickly.
LisPi
GNU social JP is a social network, courtesy of GNU social JP管理人. It runs on GNU social, version 2.0.2-dev, available under the GNU Affero General Public License.
All GNU social JP content and data are available under the Creative Commons Attribution 3.0 license.