I have been running Matrix server (again after several years), and it took 3GB of RAM just for a single user.
I try to optimize it by separating the tasks into several workers and implementing cache localityhttps://en.wikipedia.org/wiki/Locality_of_reference. Managed to reduced RAM usage down to 1GB while increasing cache size (from 10K to 20K).
I've also deployed the Matrix sliding sync, so any Matrix Client with sliding sync protocol can use the server. Pegelinux.top users can request access to it.
Add comment