1) I think that the biggest difference between how you use KV/NoSQL databases is that your objects tends to be much bigger than a "normal" SQL row. Each object can be a parent entity, plus many sub-objects and sub-arrays of information. In a relational model, that data would have to go in other tables that you'd have to join etc. I believe that this model makes things much easier for us, as all you need to debug a single failure usually is stored in a single object, instead of spread out across the entire database.
2) Very true. We're using other tools (like PostHog) for collecting and analysing user behaviour. At a bigger scale, we'd probably ship of event data to something like BigQuery for analysis.
3) We don't have QA, but for testing, objects are created in code just like any other object, and unit tests (aka most of the tests), don't need the database at all. :-)
2) Very true. We're using other tools (like PostHog) for collecting and analysing user behaviour. At a bigger scale, we'd probably ship of event data to something like BigQuery for analysis.
3) We don't have QA, but for testing, objects are created in code just like any other object, and unit tests (aka most of the tests), don't need the database at all. :-)