Apache HBase

From Bauman National Library
This page was last modified on 24 June 2016, at 13:59.
Apache HBase
fraimed
Developer(s) Apache Software Foundation
Repository {{#property:P1324}}
Written in Java
Operating system UNIX-like, Windows
Type Distributed DB
Website apache.org
Apache Hbase - it is open source, non-relational, distributed database similar to the creation of Google's BigTable and writing in Java. It was developed as part of the project Hadoop (a member of the Apache Software Foundation) and runs on a cluster of HDFS (Hadoop Distributed Filesystem), providing similar capabilities to BigTable. That is, it provides a fail-safe way to store large amounts of sparse data.

Description

  • HBase built on Apache Hadoop platform
  • To store the data used HDFS
  • Map Reduce processes may be used to download large amounts of data
  • At Reduce Loading data into a table
  • Reduce the process is carried out in the relevant region server - it is solely a local data record
  • Paradigm BigTable, developing under the Apache Hadoop, implements HBase.
  • Data is stored in the Hadoop Distributed File System,


Performance

  • 7 server cluster (16Gb RAM, 8x core CPU, 10K RPM HD);
  • Table 3 billion rows, by columns 1 to 5;
  • The size of each row - 300 bytes;
  • 300 concurrent requests;

Medium: 18ms - reading, 8ms - record

Disadvantages

  • Much more than the average (around 300ms) employs about 1% per cent of requests
  • Ability to index only a single field (row key)
  • Instability: in the latest version of the most productive possible data loss

= Disadvantages BigTable approach

  • Unstructured
  • Nonrelational

Links

1. habrahabr.ru
2.lib.custis.ru