Introduction to Hbase

Overview

Hbase is the open source implementation of Google’s Big Table database, which is where Google stores data for, for example, Google Earth and web index data.

HBase is a structured noSQL database that rides atop Hadoop. You can store Hbase data in the HDFS (Hadoop Distributed File System). And you can also store HBase data in Amazon S3, which has an entirely different architecture. HBase is structured because it has the row-and-column structure of an RDBMS, like Oracle. But it a column-oriented database and not a row-oriented one, which we explain below. And it is noSQL because you cannot use SQL to build relations between tables as you can with a RDBMS, like JOIN.

What HBase does is provide random access to big data. Hadoop does not: it is a batch system that only writes files but does not update them.

Like Cassandra, HBase stores data in a memory table and then flushes it to storage in massive writes to disk. So that gives it the high throughput needed for big data applications.

(This article is part of our Hadoop Guide. Use the right-hand menu to navigate.)

Architecture

HBase has a master-slave architecture. The master server coordinates region servers. The region servers which are responsible for writing, reading, and other data operations.

Apache Zookeeper, which is installed with Hadoop, handles configuration and distributed operations