Spark streaming : deserializaton time large with Random Forest

Question

I have a spark program, processing a stream of data from Kafka by an ML model (ensemble of two random forest). My overall RF is updated every few batches, adding (in one forest)