A Pure Aggregation Cube is a Cube that has no formulas and no dependencies. Pure Aggregation Cube is by far the best performing aggregation engine. This is because it aggregates using only the Detail elements that reside in the database. So in a Pure Aggregation Model, we can aggregate and calculate High level aggregates very quickly by simply looking at the database and knowing what values actually exists.
When building your model, using a Pure Aggregation Cube is the fastest of all models.
In the example below, the values 100 for Canada and 200 for USA sum up to the value 300 for the parent member North America.
Scrolling further in the Slice we can see that there are no values for the Members Argentina and Colombia.
Since there is no value for Argentina and Colombia, there is nothing saved in the database and nothing adding up to the Aggregate for South America. In other words, this zero value merely represents a null or missing value. Only actual typed-in numerical data elements (that are non-zero) are actually saved inside the database. This is unlike a transactional system where you can have a transactional record that has a zero value. In this Cube, every time the user changes or enters a value of zero or doesn't have a value to certain intersections, this zero actually means an empty value and therefore does not have to be calculated. This is because, when we store data in the in-memory database, we make sure to optimize it, so that the points that actually exist are easily calculated and summed into the total.
In computing for the value 300 in North America (as first pictured), it needs that aggregation engine to say that there are values for Canada and USA. Since there are no there are no other values for the other Detail members, then the Aggregate value for the Member World would total 300 also (The Detail members with value in the example being Canada and USA only). It does not need to sum up or 'aggregate' Argentina and Colombia because given the internal storage mechanism, there are only values for Canada and USA that exists. This works along all axis of the Cube, all Dimensions of the Cube along each axis.