Once upon a time, I was using postgres for OLTP and OLAP purposes combined with in-database transforms using TimescaleDB. I had a schema for optimized ingestion and then several aggregate views which produced a bunch of purpose-specific "materialized" tables for efficient analysis based on the ingestion tables.
Timescale had a nice way of abstracting away the cost of updating these views without putting too much load on ingestion (processing multiple TBs of data a time in a single instance with about 500Gb of data churn daily).
Timescale had a nice way of abstracting away the cost of updating these views without putting too much load on ingestion (processing multiple TBs of data a time in a single instance with about 500Gb of data churn daily).