Whatever approach you take, someone or something has to build a model of the data for end users to slice and dice.
The bit you’re missing is that a “cube” usually refers to a separate data store and compute engine to the data warehouse, queried via a different language such as MDX, typically meaning a proprietary tool is now required to get access. Usually this is also a subset of the data i.e. the cube designer has to decide what data is not included in the cube.
The people saying they don’t see the value of cubes are essentially saying they prefer multidimensional analysis on directly on top of a different stack: a relational, OLAP, columnar data warehouse, queried via SQL, with access to all the data.
It looks like the article is written by a vendor of such tools and hosted on their website, and others are mentioned throughout the thread.
The bit you’re missing is that a “cube” usually refers to a separate data store and compute engine to the data warehouse, queried via a different language such as MDX, typically meaning a proprietary tool is now required to get access. Usually this is also a subset of the data i.e. the cube designer has to decide what data is not included in the cube.
The people saying they don’t see the value of cubes are essentially saying they prefer multidimensional analysis on directly on top of a different stack: a relational, OLAP, columnar data warehouse, queried via SQL, with access to all the data.
It looks like the article is written by a vendor of such tools and hosted on their website, and others are mentioned throughout the thread.