I am looking for a way to do semi-additive aggregations on a non-time dimension. The model I am working with is similar to a snapshot inventory warehouse. Specifically I am analyzing processes running on a compute farm. I have one special dimension, processes, and several other regular dimensions. The problem is I would like to sum facts, such as memory consumption, on all the regular dimensions, but last child of each process.
If a processes was seen using 1GB of ram at one point and later seen using 2GB, saying the processes is using 3GB of ram is silly, but two processes each using a gig of ram I'd like to how that as 2GB.