Designing Attribute Dimensions
Essbase provides multiple ways to design attribute information into a database. Most often, defining characteristics of the data through attribute dimensions and their members is the best approach. The following sections discuss when to use attribute dimensions, when to use other features, and how to optimize performance when using attributes.
Using Attribute Dimensions
For the most flexibility and functionality, use attribute dimensions to define attribute data. Using attribute dimensions provides the following features:
-
Sophisticated, flexible data retrieval
You can view attribute data only when you want to; you can create meaningful summaries through crosstabs; and, using type-based comparisons, you can selectively view only the data that you want to see.
-
Additional calculation functionality
Not only can you perform calculations on the names of members of attribute dimensions to define members of standard dimensions, you can also access five types of consolidations of attribute data—sums, counts, averages, minimums, and maximums.
-
Economy and simplicity
Because attribute dimensions are sparse, Dynamic Calc, they are not stored as data. Compared to using shared members, outlines using attribute dimensions contain fewer members and are easier to read.
Using Alternative Design Approaches
In some situations, consider one of the following approaches:
-
Hybrid mode. If you implement hybrid mode, you can use non-aggregating attributes if needed, and you can set a custom solve order instead of using two-pass calculation. See Non-Aggregating Attributes. For more information about hybrid mode, see Adopt Hybrid Mode for Fast Analytic Processing.
-
UDAs. Although UDAs provide less flexibility than attributes, you can use them to group and retrieve data based on its characteristics. See Comparing Attributes and UDAs.
-
Shared members. For example, to include a seasonal analysis in the Year dimension, repeat the months as shared members under the appropriate season; Winter: Jan (shared member), Feb (shared member), and so on. A major disadvantage of using shared members is that the outline becomes large if the categories repeat many members.
-
Standard dimensions and members. Additional standard dimensions provide flexibility but add storage requirements and complexity to a database. For guidelines on evaluating the impact of additional dimensions, see Analyzing and Planning.
The table below describes situations in which you might consider an alternative approach to managing attribute data in a database.
Table 7-6 Considering Alternatives to Attribute Dimensions
| Situation | Alternative |
|---|---|
|
Analyze attributes of dense dimensions |
UDAs or shared members |
|
Perform batch calculation of data |
Shared members or members of separate, standard dimensions |
|
Define the name of a member of an attribute dimension as a value that results from a formula |
Shared members or members of separate, standard dimensions |
|
Define attributes that vary over time |
Members of separate, standard dimensions. For example, to track product maintenance costs over time, the age of the product at the time of maintenance is important. However, using the attribute feature, you could associate only one age with the product. You need multiple members in a separate dimension for each time period that you want to track. |
|
Minimize retrieval time with large numbers of base-dimension members |
Batch calculation with shared members or members of separate, standard dimensions. |
|
Perform cross-dimensional analysis with low performance impact |
Non-aggregating attributes |
Optimizing Outline Performance
Outline layout and content can affect attribute calculation and query performance. For general outline design guidelines, see Designing an Outline to Optimize Performance.
To optimize attribute query performance, consider the following design tips:
-
Ensure that attribute dimensions are the only sparse Dynamic Calc dimensions in the outline.
-
Locate sparse dimensions after dense dimensions in the outline. Place the most-queried dimensions at the beginning of the sparse dimensions and attribute dimensions at the end of the outline. In most situations, base dimensions are queried most.