Details
-
Sub-task
-
Status: Closed
-
Major
-
Resolution: Fixed
-
None
-
None
-
None
-
Sprint 42, Sprint 43, Sprint 44, Sprint 45, Sprint 46
Description
This idea is to write a new map-reduce for calculating cube: split the data set to several pieces (feed to mapper); each mapper calculate a cube (with all cuboids) in memory for the given data split; Then send all the cuboids which have the same cuboid id to a reducer; The reducer will merge those cuboids into one big cuboid; Finally write the cuboid into HFile;