For the kinds of processing both Spark and MemSQL do (e.g. join operation) is Spark faster than MemSQL?