Check out Presto, it is a Facebook open source product and designed exactly for your use case, it doesn't use map-reduce but can read Hive catalogs directly and

Check out Presto, it is a Facebook open source product and designed exactly for your use case, it doesn't use map-reduce but can read Hive catalogs directly and it is super fast. We use ORC as our file format but Parquet would work as well.