Used to do this by replacing slow queries with a dozen joins with stored procedures,insert/update/delete triggers, materialized views.
The mindset was correctness and validity of a detailed and interrelated collection of known facts, at rest on disk.
Merging 2 of these, say when an insurance Corp buys a competitor, took detailed and painstaking effort by people with both the domain and systems knowledge.
It's very different now,with Json doc stores, where in 5 years time noone will know what timezone that happened in, or if this person is that person with different name because of lossy utf8->ASCII .
The mindset was correctness and validity of a detailed and interrelated collection of known facts, at rest on disk.
Merging 2 of these, say when an insurance Corp buys a competitor, took detailed and painstaking effort by people with both the domain and systems knowledge.
It's very different now,with Json doc stores, where in 5 years time noone will know what timezone that happened in, or if this person is that person with different name because of lossy utf8->ASCII .