Comment

Comment on What features are missing from piefed, or, why aren't we reccommending piefed instead of lemmy?

While theoretically true, the main bottleneck with Lemmy seems to be the database performance, so with both projects depending on PostgreSQL for that, I somewhat doubt that Piefed being written in Python will have much noticeable effect in reality.

source

Sort:hotnew top

nickwitha_k@lemmy.sdf.org ⁨1⁩ ⁨day⁩ ago

the main bottleneck with Lemmy seems to be the database performance, so with both projects depending on PostgreSQL

Postgres being a bottleneck is a first for me. Not saying it’s not possible, just… It’s postgres. Wondering if it’s more an issue with ORM, etc.

source
msage@programming.dev ⁨1⁩ ⁨day⁩ ago
Postgres is so quick if you know how to use it…

source
- nickwitha_k@lemmy.sdf.org ⁨1⁩ ⁨day⁩ ago
  You don’t even need to know how to use it very well, in my experience.
  
  source
  - msage@programming.dev ⁨1⁩ ⁨day⁩ ago
    Really depends on many factors. If you have everything in RAM, almost nothing matters.
    
    If your dataset outgrows the capacity, various things start to matter, based on your workload. Random reads need to have good indices (also writes with unique columns), OLAPs benefit from work_mem, >100M rows will need good partitioning, OLTP may even need some custom solutions if you need to keep a long history, but not for every transaction.
    
    But even with >B of rows, Postgres can handle it with relative ease, if you know what you’re doing. Usually even on a hardware you would consider absolutely inadequate (last year I migrated our company DB from MySQL to Postgres, and with even more data and more complex workflows we downsized our RAM by more than half).
    
    source