Speeding up queries using BI mode

Queries of very large data sets with a number of distinct values often take too long to return results. If you can accept approximate results, using DataSketch algorithms can save significant time. BI mode integrated with DataSketches accelerates query execution while decreasing resource utilization. BI mode can rewrite a wide range of SQL analytical queries, including COUNT(DISTINCT), CUME_DIST, RANK, and NTILE.

When you enable BI mode to use Apache DataSketches approximations, a materialized view automatically calls DataSketches for some types of operations. You can pre-compute data sketches using materialized views, and then rely on Unified Analytics BI mode to rewrite algorithms to answer queries directly from those materialized views. This accelerates query execution by orders of magnitude without any change to your original queries.