At a glance
| Data model | Wide-column: rows by row key, column families, versioned cells |
| Query language | Java API, REST and Thrift; SQL through Apache Phoenix |
| Consistency model | Strongly consistent per row |
| Replication | Data durability through HDFS replication; cluster-to-cluster replication available |
| Partitioning or sharding | Automatic region splits by row-key range |
| Transactions | Atomic single-row operations; no multi-row ACID transactions in HBase itself |
| Deployment and cloud availability | Self-managed on Hadoop clusters; offered inside cloud providers' managed Hadoop services |
| Licensing | Apache License 2.0, free |
| Release lines | 3.0.0, 2.6.7, 2.5.16 (stable) |
HBase makes sense if you already run Hadoop and need fast random access to very large tables stored on HDFS, with strong consistency per row. It is free and mature, but it depends on HDFS and coordination services, has no SQL of its own (Apache Phoenix fills that gap), and outside the Hadoop world most new projects choose a lighter wide-column or managed option.
Overview
Apache HBase is the Hadoop project's distributed, column-family database, modelled on Google's Bigtable. Tables hold rows sorted by row key, with column families grouping related columns; the project describes it as hosting tables with billions of rows and millions of columns on commodity clusters, with random, real-time read and write access and strictly consistent operations. Data is stored in HDFS, and HBase splits tables into regions that are served by RegionServers and moved automatically on failure.
The project currently publishes three lines on its downloads page: 3.0.0 (5 August 2026), 2.6.7 and 2.5.16 (both 1 October 2026), with 2.5.16 labelled the stable release. HBase has a Java client and REST and Thrift gateways, but no SQL interface; Apache Phoenix, a separate Apache project, adds SQL and secondary indexes on top.
Key features
- Bigtable-style wide-column tables sorted by row key, with column families
- Strictly consistent reads and writes on a row
- Automatic sharding of tables into regions and failover between RegionServers
- Storage on Hadoop HDFS, with MapReduce and Spark integration for batch processing
- Java client plus REST and Thrift gateways
- Server-side filters, Bloom filters and coprocessors for logic close to the data
Integrations and ecosystem
Pros and cons
Pros
- Permissive Apache 2.0 licence with no editions or limits
- Handles very large sparse tables with consistent, low-latency row access
- Fits directly into existing Hadoop, HDFS and Spark estates
- Long-running project with several maintained release lines
Cons
- Requires HDFS for production storage, which is heavy to run outside a Hadoop estate
- No built-in SQL; ad-hoc queries need Phoenix, Spark or other tools
- Only single-row atomicity; row-key design determines performance
- Needs an Apache ZooKeeper quorum alongside HDFS, per the reference guide
Pricing
Apache HBase is free open source software under the Apache License 2.0, with no paid edition from the project. Costs come from the Hadoop cluster you run it on, or from a cloud provider's managed Hadoop service and its own pricing.
Pricing checked on the vendor's official page on 7 November 2026. Prices change; confirm before buying.
Who it is best for
Organisations with an existing Hadoop platform that need real-time random access to very large tables stored on HDFS.
Typical use cases
- Serving layer for very large datasets processed by Spark or MapReduce on HDFS
- Time-ordered event, log or metric storage keyed by entity and timestamp
- Large sparse tables such as user attributes or web crawl data
- Random lookups into data lakes on Hadoop
Who should look elsewhere
- Teams without Hadoop who want a simpler wide-column database: see Apache Cassandra or Google Cloud Bigtable.
- Workloads needing relational queries and joins: see SQL vs NoSQL.
Verdict
Apache HBase is still a solid, actively released choice inside Hadoop estates, with strong row consistency and scale to billions of rows. Outside Hadoop its operational footprint is hard to justify, so compare it with Cassandra or a managed wide-column service before starting something new.
Frequently asked questions
Is Apache HBase still maintained?
Yes. The downloads page lists 3.0.0 (5 August 2026) and 2.6.7 and 2.5.16 (both 1 October 2026), with 2.5.16 marked as the stable release.
Does HBase support SQL?
Not natively. HBase offers a Java client and REST and Thrift gateways. Apache Phoenix, a separate Apache project, provides a SQL layer with JDBC on top of HBase.
Do I need Hadoop to run HBase?
HBase is built to store its data in HDFS and is normally deployed on a Hadoop cluster. A standalone mode on the local filesystem exists for development, but production deployments use HDFS.
Sources
- Apache HBase home
- Apache HBase downloads
- Apache HBase reference guide
- apache/hbase repository and licence (GitHub)
- Apache Phoenix (SQL over HBase)
- Apache HBase documentation
Checked November 2026.
Listed in: NoSQL & Distributed Databases.
How we research reviews and what star ratings mean: our editorial method.