Fig. 3
From: Federated and reusable processing of Earth observation data
Illustration of different graph splitting strategies. Left: openEO process graph using two EO data sources, each loaded with a load_collection process, but one is only available on backend 1 and the other is only available on backend 2. Some more processing steps follow after that, such as merge_cubes, which merges its input data into a single data cube. This necessitates data transfer through cross-backend processing, and requires splitting the process graph in one of the four indicated split options A, B, C and D here. Middle: splitting the original process graph at split point A as discussed in Fig. 2 resulting in two new, stand-alone process graphs. Right: outcome of splitting at the split point D, which is deeper into the graph.
