Data Warehousing Using Hadoop

Abstract

The Hadoop platform supports several data warehousing solutions, including Apache Hive, Impala, and Shark. These solutions are conceptually similar to relational databases at much larger scale but differ in their implementation and usage model. Relational databases are often used in transactional systems in which single row inserts, updates, and deletes must be executed atomically. Efficient indexing and referential integrity with primary/foreign keys allow modern relational databases to find records quickly and guarantee that all data satisfies a strict schema. Relational databases try to avoid full table scans whenever possible because I/O bandwidth is limited and tends to be the bottleneck in these systems.