Overview of the most notable new features (in my opinion):

- Improved Memory Management and Configuration

- Reworked job submission (allowing things like Zeppelin or other interactive notebooks)

- Native Kubernetes integration (there's now multiple ways to submit Flink to K8s)

- Hive integration: Use your Hive tables, UDFs etc. in Flink SQL

- Full TPC-DS coverage for batch: this basically means Flink supports a lot of SQL syntax

- PyFlink supports user defined functions (based on Apache Beam's portability framework). This makes Flink's Python support really strong

... there are many more features in the announcement (1200 tickets were resolved)