Scalable

Fully distributed algorithms. GraphFrames scales as well as your Spark Cluster is scaling

Property Graphs Model

Support flexible definition of property graphs with groups and support graphs with multiple kind of vertices and edges

Feature reach

A lot of ready to use graph algorithms, like PageRank and Label Propagation

Mature

Project is old enough already and used in a lot of places as part of another libraries and platforms or as a tool for graph processing by end users

PySpark support

Not only scala, but also Python with a full support of Spark Connect protocol