Vertica Live Aggregate Projection
Featuring: Nga Tran, Software Engineer at HP Vertica
Description: Live Aggregate Projections (LAPs) are a new type of projection, introduced in HP Vertica 7.1, that contain one or more columns of data that have been aggregated from a table. The data in a LAP are aggregated at load time, thus querying it is several times faster than applying the aggregation directly on the table.
Unlike other databases which do not support incremental maintenance of non-distributive aggregate functions (e.g. MIN and MAX) for updates and deletes, incremental maintenance is possible in Vertica because the data in LAPs are partially aggregated.
This talk focuses on the implementation of Vertica LAPs, for which the current supported aggregate functions are (distributive) SUM, COUNT and (non-distributive) MIN, MAX, and TOP-K. By aggregating data per load on each partition, Vertica allows incremental maintenance on both distributive and non-distributive aggregation while allowing data to be deleted.