Apache Phoenix provides access to data that's stored in HBase which is a column oriented database. HBase doesn't natively support SQL or really any query language. The HBase shell has a scan command for accessing data but I find it extremely awkward to use. Phoenix "puts the SQL back in NoSQL" so that means if you know SQL you can query data in HBase using a language you know.

While there are multiple ways to access data in HBase through SQL including Apache Hive, Impala and Drill, I think Apache Phoenix has a better design than any of the alternatives. Unlike most relational databases that require schema design before you can add data, HBase is essentially schemaless. However all of these tools require a schema, so that means in order to use them you have to fist map the fields in HBase (at least the ones you want to use) to a metadata layer. Hive/Impala, Drill and Phoenix each do this in their own unique way, but I think the way you do this in Phoenix is the most logical.

Also in my experience Phoenix provides better throughput than any of the alternatives because it converts SQL queries into HBase scans. Review collected by and hosted on G2.com.

You do have to create the Phoenix schema manually.

Unlike the alternatives Phoenix can only access data stored in HBase, so if you want to combine data from different data sources, such as CSV or JSON files you need to use something else. Review collected by and hosted on G2.com.