1. Configuration Reference for Oracle Essbase
  2. Essbase Configuration Settings
  3. Config Settings List
  4. ASOSAMPLESIZEPERCENT

ASOSAMPLESIZEPERCENT

This Essbase configuration setting specifies the number of cells sampled from the input-level data. The sampled data is used to estimate the size of aggregate views.

Larger sample sizes enable Essbase to make increasingly accurate estimates of average view sizes. View selection using a larger sample size enables Essbase to more closely meet the stop size.

Sample sizes are specified as a percentage of input-level data.

Syntax

ASOSAMPLESIZEPERCENT [appname [dbname]] n

  • appname—Optional. Application for which sampled data is to be set.

    If you specify a value for appname and do not specify a value for dbname, the setting applies to all cubes in the specified application.

    To enable the setting for a specific cube, you must specify an application and cube.

    If you do not specify an application, you cannot specify a cube, and the setting applies to all applications and cubes on Essbase Server.

  • dbname—Optional. Specifies the cube, in the application specified by appname, for which sampled data is to be set.

    If you specify a value for dbname but do not specify a value for appname, your specification is ignored.

  • n—A value ranging from 0.0 to 100.0, representing a percentage of input-level cells that are to be used for the aggregate storage cell sample. To specify 0.5% for the sample size, enter 0.5, not 0.005. You do not need to divide by 100.

    To calculate the number of sample cells, multiply the number of input-level cells by the percentage specified in n. The default, and minimum, sample size is 1 million (1,000,000) cells.

    Note:

    For databases that have 1 million or more cells, if the percentage specified results in a sample size of fewer than 1 million cells, the setting is ignored and Essbase uses 1 million cells. For databases that have fewer than 1 million cells, the sample size is the same size as the cube.

Performance Impact

Estimates using larger sample sizes take longer to complete, which may have a significant performance impact on view selection. The recommendation for a cube with more than 1 billion input-level cells is to start with a small setting such as 0.1 (meaning 0.1%). Slowly increase this setting until the preferred trade-off between view selection performance and accuracy is reached. The optimal setting for a cube with more than 1 billion cells will probably be less than 3%. See Improving Performance When Building Aggregate Views on Aggregate Storage Databases.

To gauge the accuracy of view size estimates for aggregate views that have been built, use the following MaxL query database statement: 

query database appname.dbname list existing_views

Compare the values in the columns named size_ratio_estimate and size_ratio_actual. The accuracy of each view size estimate differs for each aggregate view.

Example

ASOSAMPLESIZEPERCENT ASOsamp.Sample 1