I care so much less about the data model. Document or Relational: I can see different apps preferring one or the other. Two things are interesting to me: 1. W

I care so much less about the data model. Document or Relational: I can see different apps preferring one or the other. Two things are interesting to me: 1. What's the consistency model? Is NoSQL really NoTxns? Whether you need transactions or not probably affects your decision more than tables vs docs. Similarly if you still want some partial functionality in the face of larger failures, maybe forgoing consistency is worth it. (Note that some NoSQL systems support strong consistency and some relational systems don't). 2. Is your query language as powerful as SQL? First, SQL is not a great language, but it's probably still better than whatever your NoSQL store is using, once you move beyond CRUD and very simple filtering. Declarative queries with an optimizer is the right choice for most applications. The expressiveness once you get to "group by", joins and subqueries is difficult to replicate in a programatic query language. You may not like SQL. That's fair. But there's a reason it seemingly can't be kept down. See CQL, Impala, SparkSQL, F1/Spanner, Presto, etc... Couchbase has been the most recent "oops" with their N1QL introduction and near complete messaging reversal (on this issue).