I'm afraid most NoSQL usage forces programmers to become, for all practical purposes, expert DBAs, or performance would be bad. Using SQL requires competence t

I'm afraid most NoSQL usage forces programmers to become, for all practical purposes, expert DBAs, or performance would be bad. Using SQL requires competence too, but learning nonstandard exotic idiosyncrasies of a certain configuration of a certain bleeding edge system (the blood is yours) is much less useful than principled and somewhat standard techniques that can be readily adapted to any good RDBMS. For example, there's a substantial difference between improving performance of a slow query by trial and error, rewriting application code to figure out what the current version of a bug-ridden interpreter/optimizer likes, and improving performance of a slow query by ENGINEERING, creating an index according to the reports of analysis tools, back-of-the-envelope estimates of size and performance and true understanding of different index types.