I made a post on asklemmy@lemmy.ml but it was not getting any responses so decided to ask here where people have experience in hosting lemmy instances.
I am not sure how cross-posting works here so I’m just going to paste the link to the original post https://lemm.ee/post/24869.
The main query I have is how is data partitioned over instances when I’m participating in communities in other instances.
You can post your answers here or there.
Looked over your other post,
Your instance A replicates or really caches the content from C@B indefinitely, if you are subscribed to it (or if anyone on your instance is subscribed). When you make a post on C@B your instance transmits that post data to B. Other instances subscribed to C@B will get your post when they receive cache updates. As long as the instances aren’t overloaded it is a pretty snappy process in my experience.
Pictures are not replicated but hosted on the instance of the user posting them. Post/Comment data is cached. Your instance will store a cache of all the subscribed communities. Even with a large number of subscriptions all that data is just text and gets filed into the database without consuming much space at all.
Edit: just some things I’ve read while setting up my instance. Pretty sure that’s accurate.
Thanks! That clears things up a bit.
So all text and metadata is kind of replicated but “heavier” stuff only resides on the original instance.
I’m assuming more the number of communities, more the instance has to periodically probe and retrieve cache updates. Hopefully these are just deltas, so the updates are small.
However, for large instances, these periodic updates can become resource-hungry activities
Hey, accidentally deleted my post instead of editing, oops.
It will have more work to do to keep a larger number of communties in sync however that part is actually pretty efficient. It sends just the bare minimum data in a json like format. If you have an instance you can watch the log and see them ticking along, getting and posting.
The majority of system load is from the UI. Database query’s to build a users home page and feeds. At least all those query’s will hit the local instances cache and spread out the load.