Integrate Presto with HBase

[from]http://labs.gree.jp/blog/2014/12/12838/

 

Phoenix

In GREE, we have a small HBase cluster, which holds data and metrics for OLAP. When we introduced Presto, we started to look for a way to integrate Presto with HBase. The solution we found is to leverage Apache Phoenix (relational database layer over HBase). Phoenix helps to create relational schemas against existing HBase tables. By building a custom connector talks to Phoenix through its JDBC interface, we are able to query HBase data directly from Presto. Furthermore, the range predicate pushdown feature works perfectly, as it pushes down the expressions specified in WHERE clause into Phoenix, and Phoenix translates them into native HBase API calls, thus only the minimum chunk of data will be fetched back to Presto workers.