In my brief review of this “Comparison” (Overview Article), it appears that Apache Beam) does overtake Apache Spark with regards to programming-model. Lines of code are cut in half in most cases while accomplishing the same tasks. More diving is required of course, but looks like something to definitely watch out for.
A little background:
Apache Beam (formerly known as Dataflow) resonated from a unification of both batch and stream processing model.
Beam is modeled after the following technologies:
and is directly powered by FlumeJava and Millwheel.
Additional interesting references: