this post was submitted on 12 Jun 2023
142 points (98.6% liked)
Lemmy.World Announcements
29104 readers
11 users here now
This Community is intended for posts about the Lemmy.world server by the admins.
Follow us for server news π
Outages π₯
https://status.lemmy.world
For support with issues at Lemmy.world, go to the Lemmy.world Support community.
Support e-mail
Any support requests are best sent to [email protected] e-mail.
Report contact
- DM https://lemmy.world/u/lwreport
- Email [email protected] (PGP Supported)
Donations π
If you would like to make a donation to support the cost of running this platform, please do so at the following donation URLs.
If you can, please use / switch to Ko-Fi, it has the lowest fees for us
Join the team
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Sorry, let me clarify. My point was that I hope Ruud doesnβt have to start shelling out thousands of dollars because everyone decides to register there and he needs to keep increasing capacity.
Thing is, lemmy doesn't support clustering/horizontal scaling. So there are limits to how much increasing you can do. You can beef up with a database cluster, add a separate reverse proxy, and increase the specs of the hardware lemmy is running on (but hardware can't be expanded limitless), but that's about it. Once you hit the limit of what a single instance of the lemmy software can handle, you cannot scale anymore. Pretty sure you will hit the limit long before you reach thousands of dollars.
Hopefully by then, Lemmy's source is augmented to support HA/horizontal scaling
Yeah. But horizontal scaling (well horizontal scaling in a system like this where you need clustering so the instances talk to each other) is hard. And I think there are a lot of other things that need to be polished, added and worked on before that. It would probably also need somebody with knowledge of clustering to start contributing. I think step 1 needs to be that the dev team needs more help properly tuning the database use. The database is very inefficient, and they lack the skill to improve it:
So getting help improving the database is probably the #1 thing that can be done to deal with the scaling problem.
I fully agree - there's no excuse why the DB should be falling over when I've seen a single postgres instance (with a read replica, granted) handle >1M users just fine.
Unfortunately my SQL skills haven't improved since my DB class in university, so I won't be much help. I'll be keeping an eye on the repo of course, and I can give some consulting/guidance or even open some PRs myself when they decide to implement horizontal scaling.
Yeah. I'm in the same boat. My SQL skills aren't impressive either since there are other people at work that handle optimization. Haven't used rust either (yet) so cannot really contribute there either. Though I'm considering potentially starting work on a cross platform mobile app. I haven't worked with mobile apps for a good six or seven years, so I feel like it's high time I get back up to speed. (But knowing me, I'll end up making something half finished and the start procrastinating)
Yeah, I think a single postgres cluster with read replicas should be able to handle the needs of a single instance just fine. You can then horizontally scale the backend and frontend to keep up using containers.
Ideally, that instance can scale up to a few million users, and then federation can provide the horizontal scaling that takes the lemmyverse up to Reddit scale.
The backend just needs to handle databases better. Adding support for read replicas, making it more efficient, etc.
Not sure how well pict-rs scales, but that's probably pretty light already, vertical scaling might be good enough that it'll always be limited by the DB.
But yeah, I guess the worst case scenario is that Postgres doesn't scale enough and we need to switch to something like cockroachdb. Or go for snowflake uuids and noSQL like Twitter did back in the day.
I believe that's mostly because of websockets. The devs are changing the UIs from using websockets to just normal html rest api calls. That should help with a lot of load, since you will have to manually hit refresh to reload the site (like reddit), and will allow scaling horizontally since each html request doesn't need to run on the same server as the previous request.
Getting rid of websockets would help a lot. But you still might not be able to have standalone nodes. You might still need a cluster of nodes with a master and slaves due to the federated nature of lemmy. Such that only one node at a time can handles federation events with other servers. I don't know enough about the protocol to know if that is the case or not. Just as an example I'm thinking of situations where one node gets a federation event for example for a post, then a different node gets a federation event with some sort of change to that post, and handles it faster than the first node. That event would then fail because the post hasn't been created yet.
Interesting. I didn't think of that. Maybe some sort of queue would help, but yeah not sure. Maybe the protocol can handle that already. I'll have to read through it at some point. :-)
Oh yeah, definitely.