Thank you for your answers, and sorry for my lack of understanding.
So I tried what you suggested, with/without unpersisting and with .cache()
(also persist(StorageLevel.MEMORY_AND_DISK) but this is not allowed for msg
because you can't change the Storage level apparently) for msg, g and
newVerts,
I actually tried without unpersisting, but given the performance I tryed to
add these in order to free the memory. After your anwser I tried to remove
them again, but without any change in the execution time...
Looking at the web interface, I can see that the mapPartitions at
GraphImpl.scala:184
Hi,
I'm working on big graph analytics, and currently implementing a mean field
inference algorithm in GraphX/Spark. I start with an arbitrary graph, keep a
(sparse) probability distribution at each node implemented as a
Map[Long,Double]. At each iteration, from the current estimates of the