Replies: 1 comment 2 replies
-
From the standpoint of an electric app, the local database will only contain the subset of data that is being synched via shape subscriptions (e.g. the relevant data for the group). On the client you get the electrified schema as defined in Postgres, but performance is not tied to the backend database size or workload patterns. It is important to scale the backend database, but at the moment, partitioning or sharding would greatly depend on how transparent extensions are with regards to schema management and logical replication. We haven’t explored more for now. |
Beta Was this translation helpful? Give feedback.
2 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
One of the benefits of a pouchDB/couchDB data stack is that, for user groups with isolated data (companies, teams of users, etc), partitioning is very natural as each isolated dataset can be put in its own "TableBase" (single-table database).
A system where every user group's data accumulates in a PG table is an eventual issue without eventual partitioning.
Does electric support partitioned tables on postgres? What about sharding? e.g with the tidus extension?
Beta Was this translation helpful? Give feedback.
All reactions