PowerOLAP® is an on-line processing application. In other words, users can change data points within a model and have those values immediately calculate into new aggregates.
There are a couple of situations in PowerOLAP® that can affect what happens to the cache when data values are changed. The first instance is in a Cube that is a pure aggregation Cube. A pure aggregation Cube means that there are no formulas within that particular Cube. If a detail point is changed in a pure aggregation Cube, PowerOLAP® will simply clear those aggregates or the portion of those aggregates that are dependent on detail elements.
The second instance is when the Cube has formulas. In this situation, PowerOLAP® must clear the entire cache. This is necessary because of the variability of what users can do with the PowerOLAP formula: a formula can perform any operation anywhere in the Cube.
PowerOLAP® clears the entire cache in a Cube that has formulas when a data point is changed.
This can be a significant performance hindrance in models where this situation exists. When designing a model, keep caching in mind. For example, a model where one Cube handles data entries, a second Cube performs formula calculations at an aggregate level, and lastly reports are created using the values from the latter Cube; this represents an ideal or optimized model for caching.
Example 1: Pure Aggregation Cube (No formulas within the Cube)
In the following screenshots, the data for the detail point Canada for the Month of February has been changed. When a user refreshes or recalculates the Slice, PowerOLAP® will only clear the cache for the aggregates North America and World for the month of February, as well as the Aggregate value of 1st Quarter and Total Year. The values for the rest of the months and other aggregates will remain intact.
This can provide a significant performance improvement provided that the user is using a pure aggregation Cube model.
An example of Pure Aggregation cube that does not contain a formula
Second Example: Cube with a Formula
In the given example, if the Cube contains a formula, changing the value of Canada for the Month of February could, based on the formula, impact the values of Canada for other Months or the value of other Regions like USA for a particular Month.