The database-backed middle comes with all the advantage a database storage brings but has grown to a considerable size.
It would be great to have an additional middle implementation with some kind of custom disk-based database, this could be based on something like RocksDB or written from scratch exactly for our use case. Users could then choose whether they want the more efficient one or the more flexible PostreSQL-based one.
Ideally there would be a foreign data wrapper for PostgreSQL, so that users still can access the middle data through SQL. That would allow us to combine the advantages of both solutions.
The database-backed middle comes with all the advantage a database storage brings but has grown to a considerable size.
It would be great to have an additional middle implementation with some kind of custom disk-based database, this could be based on something like RocksDB or written from scratch exactly for our use case. Users could then choose whether they want the more efficient one or the more flexible PostreSQL-based one.
Ideally there would be a foreign data wrapper for PostgreSQL, so that users still can access the middle data through SQL. That would allow us to combine the advantages of both solutions.