Hive provides no additional capabilities to MapReduce. The programs are executed as MapReduce jobs via the interpreter. The Interpreter runs on a client machine which rurns HiveQL queries into MapReduce jobs. Framework submits those jobs onto the cluster.