what is better for single user instance, or maybe something small like under 10 users (no communities)? which is lighter on resources? how much storage should I allocate?
any alternatives to lemmy and kbin that are still somewhat similar?
what is better for single user instance, or maybe something small like under 10 users (no communities)? which is lighter on resources? how much storage should I allocate?
any alternatives to lemmy and kbin that are still somewhat similar?
While I quite enjoy the Interface of KBin over Lemmy. It seems Lemmy uses a lot less resources based on the “Admin Guides” for each service.
Lemmy: https://join-lemmy.org/docs/en/administration/administration.html
KBin: https://codeberg.org/Kbin/kbin-core/wiki#user-content-admin-guide
My Lemmy instance is currently occupying about 350MB of RAM, but you can round that up to 400MB. A lot less than the 4GB for KBin.Technically it’s a dual user instance now, since a friend wanted to join it and I said sure.
Damn, this is so much worse than good ol’ RSS for just following up stuff (which I imagine is the main argument to be made for a single-usee instance)
The arguments for are varied. I don’t have to worry about any admins making decisions on federation, I can federate (or not) however I please. I have my own space that I can do what I want with in a familiar format, and I can make my username Jamie without it being taken.
Yep, making federation decisions myself is why I want to spin up my own instance at some point, and I have spare computing resources as is already lol
fair points! But that’s a high price to pay in terms of computing resources.
I wish federation in the “fediverse” sense was as inexpensive as in the XMPP world (or at least seem to be)
High price? It’s really not very intensive at all, at low user count. Super light on storage too. I think BeeHaw (one of the biggest instances) said their whole instance was only like 25GB (a week ago)
Yeah, I mentioned on another comment that storage space is the last thing to worry about, 1GB of storage is typically a fraction of a penny.
I meant, comparatively: you can host thousands of active jabber accounts on a RPi at the same time, and that will cost you pennies in electricity a week. I know this isn’t close to an apple-to-apple comparison (different protocols, different capabilities and features, …), but what interests me when people bring-up federated protocols is how much will it actually be used that way in practice (wrt. server dimensioning vs number of users, effort to set-up and administrate, etc), and how effectively we are breaking away from the centralized internet that’s so nefarious. And sorry if this is shifting the thread away from where we started at :)
One more anecdote for y’all to pluralize into data: my instance is currently using 915Mi of storage for pictrs and 976Mi for postgres, roughly 650Mi of RAM (including postgres), and negligible CPU
It’s not that bad, I’m hosting my instance in a small corner of a friend’s server and not making any impact on it.
My other reason is that it’s the only way to know I picked an instance that isn’t going to just go away without me and take my account with it. It will be an interesting day when the first major lemmy instance goes down…
Yeah, if I stick with Lemmy (which seems highly likely) I want to put up my own private instance for similar reasons and maybe even invite friends and family that I can talk into it.
yeah for real I have been thinking the same lol. sadly a feed reader is just a reader. Would not be able to comment or post, which is the main reason I will go with a self hosted instance.
As a middle ground you could follow a Lemmy community‘s RSS feed then jump in via the link for anything you wanted to see the conversation for :)
Keep in mind as php-fpm (used by Kbin) launches multiple child processes instead of a single rust process like Lemmy. The number of child processes can be tuned down to reduce RAM usage if necessary.
Thanks for the info! what about storage capacity usage?
After almost 24 hours, coming up on 662MB of images, and 371MB for the postegres database. Though, I could see the numbers fluctuating depending on how much stuff you’re subscribed to. I’m currently subscribed to 31 communities, most of them fairly large.
I’m still kinda new to lemmy, but it sounds like every post on all your communities get sent to your instance? And maybe every comment?
That’s probably fine for now, but when happens when one of the communities you follow gets a ton of users? I imagine you’d end up having to scale your self-hosted server even though it’s just you consuming the content?
That doesn’t seem sustainable. Not knocking your idea to self-host, more concerned with the scalability of lemmy
That’s correct. I’m replying to you on jamie.moe right now, but in reality, what I’m seeing is a copy of what lemmy.world has. When someone makes a post, leaves a comment, or whatever, all of the instances that are federated and have one subscriber in the relevant community get updated with the change.
When I read this thread, I’m actually reading a copy of the thread stored on my own instance, and my instance just tells lemmy.world when I leave a comment and it will update everyone else. While it isn’t very storage efficient, it does even out the load since users on one instance only require a single update from the main instance for all of those users to interact with that content, rather than instances asking for copies of things every time a user on one instance needs it. It spreads out the load nicely and lets even underpowered hardware serve wider communities than it could otherwise.
The downside is you have to have the storage space to keep content for your user(s). Though, getting a lot of storage space isn’t expensive at all, so I would say it’s a minor concern. Lemmy uses little in terms of RAM and compute, those two factors are the biggest things that can up your cost really fast. A gigabyte of storage on a server can typically be gotten for a fraction of a penny.
Why so much for images? I thought Lemmy hotlinked those?
It stores them on a local pictrs server and pulls them from there, which, to my understanding, holds them for a week.
BeeHaw.org admin said like a week ago they were only up to 25GB.
Yeah, kbin seems to be a php8 service vs a compiled rust service from some lite poking.