Question about matrix computation

[Marcos Takahashi]

I am new on RHadoop and I would like to know about Matrix implementations.

I read about using Matrix as tuples (row, column position and value) and also splitting it to fit on each node's memory size.

Which would be the most suitable implementation? I'm implementing an ALS Collaborative Filtering, so it's very important to know what works better.