The stuff I liked most in the database releases of the last years is the In Memory functionality. Why? Actually it’s simple. Simple because it’s easy to enable. Simple because it’s solving the problem all those databases have were they combined OLTP and DWH characteristics. And…there are a lot.
Also it’s a tool that can be used for stuff there isn’t a proper solution for like how to deal with semi- and unstructured data. For example via using materialized views you can make JSON content (or XML for that matter) available in memory, the pieces in which you’re interested in, and make selecting this content lighting fast.
Also in memory might be the means to the almost utopian wet dream: a self tuning database for SQL. Ah okay, agree, we are not there yet. But it would be a thing to strive for. “Execute some SQL and it will perform with knowing only basic SQL”.
Yeah I like the “In Memory Column Store” a lot. So still more to come on this site. If not only there are cool new 12.2 features coming.