Date: Nov. 3, 2016
Time: 11:50 - 12:40
Room: Alfa 2
With geospatial dataset always growing larger, it is now common to have to handle terabytes of data. GiST indexes can be successfully used in PostGIS, but with such volumes of data, their performances tend to drop as their size grow larger than available RAM. For smaller dataset, they perform well but are big and can take a long time to build.
This has prompted us to develop a patch for PostGIS, in order to add the support for BRINs for geospatial datatypes. It has been submitted and recently reviewed, and have been committed into PostGIS.
In this presentation, we will present how BRIN indexes can be used on geospatial data. Benchmarks will be presented comparing build time, index sizes and performances in read/write queries of BRIN versus GiST indexes. We will also explain current limitations of BRIN indexes and how we can further improve them.