If you have tons of data (logs, for example), on the order or tens or more terabytes, and want the benefit of SQL in addition to MR to query and crunch that dat

If you have tons of data (logs, for example), on the order or tens or more terabytes, and want the benefit of SQL in addition to MR to query and crunch that data offline, you can use something like Hive which runs on top of Hadoop, gives you all the power and familiarity of SQL, and lets you "drop down" to MR if you need the extra power.