1. Home
  2. NoSQL Databases
  3. HBase Reviews
  4. User Review of HBase
Apache HBase: Through the Looking Glass!
Updated November 25, 2015

Apache HBase: Through the Looking Glass!

Rekha Joshi | TrustRadius Reviewer
Score 7 out of 10
Vetted Review
Verified User

Overall Satisfaction with HBase

Apache HBase was used for mastering solutions, for creating master data sets and reconciling conflicting data coming to Apache Hadoop systems.
  • Apache HBase is a widely used java based distributed NoSQL environment on Apache Hadoop.
  • While there has been growing interest and efforts in in memory computing, there are investments on Apache Hadoop (or hadoop provider variants) across domains. So that is a large market.
  • I worked on HBase for applications which needed to provide strong consistency and interact with Apache Hadoop.
  • You could encounter issues like region is not online or NotServingException or region server going down, out of memory errors.
  • As HBase works with Zookeeper, care needs to be taken it is correctly set up. Most issues pertain usually to environment setup, configuration, shared load on system or maintenance.
  • The performance across workloads when evaluated against other NoSQL variants was not best in class, this is most times okay, but can be improved.
  • If you use Apache HBase, and want to upgrade it for some features then you might need to do a compatibility check against your Apache Hadoop and Apache HBase versions, there are dependency to think about.
  • The HBase master slave becomes the single point of failure, and may not be a preferred design.It is not highly available system.
  • Last I checked it did not have well tested easy integrations with Spark, and that can help.
  • Faster data insights
  • Better customer service
These days I use Apache Cassandra more for even more scalability, good performance under different kind of workloads, and for providing highly available systems. Apache Cassandra also has connectors for Hadoop, Spark, and Solr.

Both scalability and response times are reasonable in HBase once tuned correctly. I hear latest version of HBase 2.0 is even better! Needs to be evaluated.

However it entirely depends on projects and what are we trying to solve when making this decision.

The key questions I ask when choosing NoSQL distributed database:

  • What is the application's inherent need? Does this component fit well in the design?
  • Does it provide high data security?
  • How does it assure there is no
    data loss?
  • How can I make sure it is a highly available system, and no downtime for customer?
  • Does it give me the best linear scalability?
  • What kind of tuning parameters does it allow the user to configure?
  • How does it stack up against other
    NoSQL
    variants on features, scalability, ease of use/contribute to and maturity of product?
  • What throughput can it attain under different kinds of workloads?