Is there a difference between cluster modes in this case? Can it be that Graphx would work better on single than on standart cluster or high concurrency cluster (for multiple users)? Does less concurrent cluster wourld be more efficient for graph modelling?