Working with different databases allowed me to see that different data setups require different database setups, simply for performance. The move to a microservice architecture allows to optimise this even further as it enables the engineers to choose the best possible data storage setup for a very small domain. Some solutions might not even need data storage but can live with an in memory cache storage. Other solutions might need relational functionalities requiring Postgres or MySQL and co.

Luckily, most modern development frameworks provide database angosticity allowing code to run on different database setups. This provides an easier way of migrating when the data model becomes more complex than originally anticipated. Most important remains that users should never ever have to worry about data, not about it's format, nor storage, not retention and not recovery.

Streaming architectures allow a completely different approach to data with flow architecture being the extreme where data is present and accessible at any time, without any service (except the streaming service) making it available, transforming it, processing it.