when using Spark, I would normally do one of two (k, v)
- Each point in the cube is given a key of (i,j,k). While making the extraction of subsets like planes very easy, it isn't very space efficient
- Each record is given an (i,j) key. Map functions know have to iterate, which isn't very pretty at all
I'm hoping that thunder has a better idea that what I've come up with so that I can extract planes out of the collection of series.