The requirements related to the BI Blend environment vary widely by the volume of source records together with the BI Blend settings definition.
Default Server Selection
BI Blend processes are queued across the available Stage servers. On the BI Blend workflow settings, a defined server can be assigned to dedicate all BI Blend processing.
NOTE: This should only be used for Platform Versions 7.4.4. and earlier. Platform Versions 8.0.0 and later should not have defined servers.
Learning Mode
Learning Mode occurs during the first process instance of BI Blend and the design related to the choice of Blend Unit. This mode restricts multi-threading to two threads by two Blend Units to generate predictive statistics based on the number of records generated from the BI Blend settings. Subsequent processes are optimized to multi-thread each Blend Unit. Should the Blend Unit be changed, the Learning Mode is enabled again. Additionally, if the number of aggregating dimensions is increased from the prior settings, the Learning Mode is enabled again.
-
Only two threads run
-
Default mode when BI Blend Task is run for the first time
-
Limited multi-threading is done to help ensure free memory is not exceeded
Log File Statistics
-
Blend Unit
-
Base Rows
-
Parent Factorial
-
Explosion Factor
Second Pass Processing
After a successful import in Learning Mode, the same thread can be evaluated. BI Bend processing will observe the current number of rows in each Page dimension (Blend Unit) and apply the Explosion Factor to determine if the process will exceed the amount of free memory available on the server.
If a new Blend Unit is added to the file, the calculation estimations for logging and memory usage will use an average across the Page member statistics in its calculations.
