HBase - good for UI performance from a Hadoop cluster
April 14, 2017
HBase - good for UI performance from a Hadoop cluster
Overall Satisfaction with HBase
HBase makes it possible to provide sub-second UI responsiveness when querying very large data sets. This is in contrast to something like Hive, which could take many minutes.
- Very fast query capability
- Resilient: by leveraging hdfs, hbase can handle server failure pretty well
- Very schema dependent - you have to carefully choose your schema and key strategy in order to get good distribution and performance.
- Over aggressive rebalancing - if you have to bounce your system - for example - hbase will spend quite a while trying to rebalance all the data as each server comes online.
I like using Apache Phoenix to simplify JDBC access. Without Phoenix, hbase feels a bit inaccessible for folks mostly familiar with SQL.
- Facilitates building of UI tools that provide valuable business insights
Hbase is less robust but faster.