Hi, I want to control the placement of the partitions of the Property Graph across my cluster nodes. As I understand, in order to specify the preferred locations for a partition of an RDD, one will need to create a subclass that overrides the getPreferredLocations() function. For example the ParallelCollectionRDD overrides that method to take into account the preferences.
However, Property Graph in GraphX is combination of multiple RDDs. How can I provide the preferred locations for it? Would I need to implement only a custom EdgeRDD with the getPreferredLocations() function? Since it seems that VertexRDD is partitioned according to the edge partitions. Do I need to implement something else as well? I have asked the same question on stack overflow as well: https://stackoverflow.com/questions/55346931/custom-partition-placement-graphx Thanks.