The IBM Cognos TM1 calculation engine uses FEEDERS to help manage sparsity in cubes with SKIP CHECK enabled. There may be rule calculations for certain cubes, but they are extremely small or dense and may not need SKIP CHECK,More info go through Cognos TM1 Course.
FEEDERS classify the cells in a cube which may contain a measured value based on a rule and ensure that these values are included in the calculation of the aggregation.
SKIP CHECK is a keyword entered at the top of a rule file that makes the sparse data consolidation algorithm of TM1 possible. Usually, cubes using SKIP CHECK would require FEEDERS, another keyword entered in a rule file later on. For more on SKIP CHECK, please refer to section 2.3.
Why FEEDERS for use?
OLAP cubes can be very sparsely populated as it is very important to handle data within TM1 cubes.
Examples of problems facing the construction of the TM1 Cubes are below.
Example 1- Cube Sales
Take a sales cube with the following dimensions and number of elements:
● Magasins (500)
● Geography (300)
● Items (50 thousand)
● The (6) networks
● Unit of Business (12)
● Moment (300)
● Acts (15)
Count of cube cells = 500 * 300 * 50000 * 6 * 12 * 300 * 15 =2,430,000,000,000,000 cells.
TM1 stores all data as a "double," but depending on the number of dimensions in the cube, the RAM consumed by storing each cell depends. For data storage, a 5-dimensional cube would need ~143 bytes per cell (excluding RAM needed for fed cells)
For calculations in a sparse cube, using standard dense matrix algorithms will consume large quantities of computational resources and take a considerable amount of time to complete. IBM Cognos TM1 uses a sparse data aggregation algorithm by default to boost calculation efficiency in sparse cubes. This enables IBM Cognos TM1 to very easily and efficiently conduct dimensional
aggregations or consolidations. Bear in mind that, depending on the hierarchical configuration of the dimensions used by the cube, these dimensional aggregations or consolidations are measured. They
are distinct from rule-based calculations that are used inside or across dimensions and cubes to calculate various cells.
Dimensions of the Cognos TM1 product
A product dimension containing the items Product A, Product B, Product C, and Total Products will be an example of dimensional aggregation or consolidation. Any cells with values for Product A, Product B, and Product C will be aggregated into Total Products based on the hierarchical nature of
the dimension. A rule-based calculation example would be Price * Volume = Revenue. In IBM Cognos TM1, rules are specified in the TM1 Rules Editor outside of the dimension and are then saved to a TM1 cube.
Cognos TM1 cube
However, once IBM Cognos TM1 detects that a rule has been applied to the cube (when the .rux file has been generated and saved), TM1 will automatically disable the sparse data aggregation algorithm. This is to ensure that values that include the rule calculated values are calculated in the
aggregated calculations. The values determined by the rule itself will always be "right". Performance degradation will be experienced on average by large and sparse cubes, while smaller cubes may not.
Sparse consolidation Cognos TM1 would return the right result if the sparse consolidation algorithm was turned off, but the machine would become considerably slower.
All of the above solutions are unacceptable, which resulted in the implementation of the idea of feeders. Feeders are a way of enabling the performance advantages of the sparse consolidation algorithm to continue to be leveraged by cubes containing rules, while also ensuring that measured
rule cells are not skipped when calculating consolidated cells.
Sparse cubes without consolidation and requirements for zero suppression may contain rule calculations without feeders. Also, the denser a cube, regardless of size, the less the skip check and feeders can gain from the cube.
What does SKIP CHECK mean?
SKIP CHECK is used as the sparse consolidation algorithm in the TM1 rules. It effectively overrides the default TM1 behavior with rules for cubes.
By default, what is the setting?
SKIP CHECK is ON by default before a rule file is generated that essentially switches off the skip check.
How do Cognos TM1 FEEDERS work?
Feeders, unlike guidelines, only ever apply to cells at the leaf level and never to consolidated cells.
Consolidated elements can, however, be used as a shorthand way of defining all leaf elements within the consolidation in the definition of a feeder sentence. The following occurs when defining a consolidated element in a feeder statement.
Feeding on a consolidation means that if there is a value present, all the leaf descendants of the consolidation will eat. For example, if there were four-leaf elements in the consolidated element but only two of them contained a value, then only those two would eat.
Feeding a consolidation means that it will feed all the leaf cells under the consolidation. For example, if four-leaf elements were contained in the consolidated part, then all four leaf elements would be supplied.
The essential concept for Cognos TM1 feeding
This is an essential concept since some individuals erroneously assume that it is the consolidation that feeds or is fed. If this idea is not well known, this can make debugging rules and feeders more difficult.
A single byte "flag" or "pseudo data" in a leaf cell is set when a feeder is applied to signal that it should be consolidated. This makes sure that this cell is not skipped by the sparse consolidation algorithm when computing consolidated values. A cell remains fed until either the server is restarted or the cube is unloaded.
In general, the C-level rules do not include feeders. The only exception is when a rule is applied to a combined element where none of its child elements have values.
Overfeeding and Underfeeding in Cognos TM1
It is the function of the author of the rules to ensure that measured cells are correctly fed. It is important to take care to ensure that feeders feed exactly as many cells as required, no more and no less. If a model is not properly fed, considerable side effects can occur.
Underfeeding in Cognos TM1
Underfeeding happens when it is not fed to any or all of the values being measured. This will, in most situations, contribute to incorrect results when the consolidated data in the cube is checked.
Underfeeding, since it compromises the model's dignity, must be prevented. You should start with the inverse of the rule you are feeding when writing feeders to guarantee that feeding is adequate.
Overfeeding in Cognos TM1
Overfeeding occurs when (a) cells that do not contain measured rule values are fed, or (b) when cells that are measured by a rule resulting in a zero value are fed. It is important to avoid all of these circumstances, but particularly (a). While overfeeding does not lead to incorrect values in the cube, it has a detrimental effect on the performance of the system. It wastes the time needed to feed cells that do not need feeding, and all unnecessary feeders take up memory that is also wasted.Overfeeding can cause a major memory explosion and can increase consolidated query execution time.
I hope you conclude Cognos Tm1 feeders. You can learn more through Cognos TM1 online training.