this post was submitted on 11 Jun 2023
169 points (99.4% liked)
Asklemmy
43956 readers
1583 users here now
A loosely moderated place to ask open-ended questions
If your post meets the following criteria, it's welcome here!
- Open-ended question
- Not offensive: at this point, we do not have the bandwidth to moderate overtly political discussions. Assume best intent and be excellent to each other.
- Not regarding using or support for Lemmy: context, see the list of support communities and tools for finding communities below
- Not ad nauseam inducing: please make sure it is a question that would be new to most members
- An actual topic of discussion
Looking for support?
Looking for a community?
- Lemmyverse: community search
- sub.rehab: maps old subreddits to fediverse options, marks official as such
- !lemmy411@lemmy.ca: a community for finding communities
~Icon~ ~by~ ~@Double_A@discuss.tchncs.de~
founded 5 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
(This is purely my personal opinion, of course!) In the scenario in which a few large instances dominate, the idea of the fediverse failed. One may estimate the likelyhood of success or failure given how they expect humans to behave, but in the end experiment beats theory. I think that for the fediverse to work a significant cultural shift has to occur, but I don't think that it is an impossible shift. I would like the fediverse to succeed, and so I choose to take part in the experiment.
Yes, that might cause some serious issues. The project is still in an early-development phase, and I don't understand the technical aspects well enough yet to be able to identify whether there is obviously a fundamentally invincible barrier when it comes to scalability. My optimistic hope is that the developers are able to optimize horizontal scalability fast enough to meet the demand for scale. If it turns out to be impossible to scale, then only rich enough parties would be able to have viable instances, and that could be a reason for failure.
What does ‘horizontally scalable’ mean here? I haven’t come across that before.
This is what I think, but if anyone understands it differently please correct me.
Vertical scalability refers to scaling within a single instance. More users join and they post more content, increasing the amount of disk space needed to hold that memory, network bandwidth to handle many users downloading comments and images at once, and processing power.
Horizontal scaling refers to the lemmyverse growing because of the addition of new instances. The problem in this form of scaling is due to the resources that an instance has to use due to its interactions with other instances. So, you may create a small instance without a lot of users, but the instance might still need a lot of resources if it attempts to retrieve a lot of information (posts, comments, user information, etc) from the other larger instances. For example, at some point a community in lemmy.ml might be so popular that subscribing to that community from a small instance would be too much of a burden on the smaller instance because of the amount of memory required to save the constant stream of new posts. The horizontal scaling is a problem when the lemmyverse becomes so large that a machine with only a small amount of resources is no longer able to be part of the lemmyverse because its memory gets filled up in a few hours or days.
You can summarize by thinking of vertical scaling as "make machine bigger / more powerful" with horizontal scaling as "make more machines".
Kind of like building a very large/tall building vs having multiple buildings!
I don't believe this is how it works though.
Let's say your tiny 3 person instance is connected to a big one. I believe it only pulls in content from the communities somebody from the small instance is subscribed too. Correct me if I'm wrong.
That's what they're saying.
Essentially - if someone from the small instance subscribes to a community that has a ton of data (huge post volume, images, whatever), the small instance needs to pull data over from the larger instance. At some point there may be communities that are so large small instances can't pull them in without tanking.
I wonder if there is a way to get around this? maybe smaller instances will have to be text-only?
maybe I phrased that poorly and you didn't understand what I was trying to say. The size of the bigger instance shouldn't matter at all because only data from communities is pulled, that a member of the smaller instance is subscribed to. So if the bigger instance has 1000 members or 2 million members wouldn't make a difference. The only thing relevant should be how active the communities are that members are subscribed to.
Could that be solved by caching? Can't the smaller instance avoid some duplication?
If I'm reading the protocol right, it's probably larger instances that will avoid more duplication, since:
I'm not sure I see where you see caching fitting in.
I am surprised I don't see some kind of lower resolution digest concept in the protocol (which might be what you're looking for)
That's what I've gathered, but I don't believe there's a way for instance owners to limit what's fetched - a user crafts the query and the server does the needful.
I imagine this could amount to a denial of service attack of sorts, if some high-churn communities are imported into tiny instances. How bad that could be, I have no idea - I'm speaking pretty theoretically, here. Text is tiny, after all, so it's probably not much of a concern, since most of the media is actually handled elsewhere...
I'm not a web developer. I'm sort of a sysadmin so i have some experiences maintaining machines for web apps for other people. And you are right...text will not create massive amounts of data. But a lot of tiny transactions can bring down machines surprisingly fast even if the total amount of data is relatively small.
I guess we are here to experience it first hand. I don't think anybody...not even the developers have a clear idea of how well this will scale. There is only one way to find out lol
Some things can go faster if you add more workers, some things can only go faster if you make the workers bigger or faster .
If you're tidying a garden you can get it all done more quickly, and tackle bigger gardens, by getting your friends to help. That's horizontal scaling.
If you need to get a parcel from your house to Burkina Faso the only way to do that more quickly is to use a bigger, faster machine. That's vertical scaling.
The way Lemmy is designed right now (says the op, I don't know the detail) you can only support more users by making the server bigger and more expensive, not by using lots of smaller servers.
Edit: note that Lemmy as a whole scales horizontally: more instances == more users, but each instance has to scale vertically.
I think that having few big instances is not failing, it is natural for social network (where lemmy is some representation of one) to be scale-free network, which has big hubs and buch of smaller nodes connected.
Most people would go to general instances, but artists will probably go to some art focused instabce, developer to proggraming.dev... But we will have bih hubs, there is no way around it.
Yes, you are right, I should have formulated that better.
I would expect that there would be a few big instances. What I should have said is having only a few big instances and no small instances would be a failure. It would be totally acceptable to have a few big instances and lots of smaller instances that can still interact with the fediverse. The failure would be if you have something like 20 very big instances that only interact with each other and that are inaccessible to the small instances - either because they close their federation, or because it is too resource-intensive for small instances to interact with them. In this case you end up with a centralized system again, not better and potentially worse than something like Reddit. As long as someone can spin-up their own instance if they want to and be part of the larger ecosystem, it would be a success.