Table of Contents

Search

  1. Version 10.0
  2. Version 9.6.1
  3. Version 9.6.0

Partitioned Mappings

This section describes changes to partitioned mappings in version 10.0.
Effective in version 10.0, the Data Integration Service can create a different number of threads for each mapping pipeline stage. The service determines the optimal number of threads for each pipeline stage. The number of threads created for a single pipeline stage cannot exceed the maximum parallelism value.
Previously, the Data Integration Service calculated a single actual parallelism value and used that same value for each mapping pipeline stage. The service calculated the actual parallelism value based on the maximum parallelism values and on the maximum number of partitions for all flat file, IBM DB2 for LUW, or Oracle sources ready by a mapping.
Effective in version 10.0, you can disable partitioning for a Decision or SQL transformation by clearing the
Partitionable
advanced property for the transformation. The Data Integration Service uses one thread to process the transformation, and can use multiple threads to process the remaining mapping pipeline stages. You might want to disable partitioning for these transformations because these transformations might not return the same result for each mapping run when they are processed with multiple threads.
Previously, the Decision transformation did not support partitioning. When a mapping contained a Decision transformation, the Data Integration Service did not create partitions for the entire mapping. The SQL transformation did support partitioning. You disabled partitioning for the entire mapping when this transformation needed to be processed with one thread.
Effective in version 10.0, if a mapping establishes order with a sorted relational source or a Sorter transformation, the Data Integration Service can use multiple threads to run the mapping. To maintain order in a partitioned mapping, you must specify that targets maintain the row order in the advanced properties for the Write transformation. When you configure Write transformations to maintain row order, the Data Integration Service uses a single thread to write to the target.
Previously, if a mapping included a sorted relational source, the Data Integration Service used one thread to process each mapping pipeline stage. If a mapping included a Sorter transformation, the Data Integration Service used one thread to process the Sorter transformation and all downstream mapping pipeline stages.
If you upgrade from an earlier version, all existing Write transformations are configured to maintain row order. The Data Integration Service uses a single thread to write to the target to ensure that any order established in the mapping is maintained. If an upgraded mapping does not establish an order, you can clear the
Maintain Row Order
property in the advanced properties for a Write transformation so that the Data Integration Service can use multiple threads to write to the target.
Effective in version 10.0, you can disable partitioning for a Java transformation by clearing the
Partitionable
advanced property for the transformation. The Data Integration Service uses one thread to process the transformation, and can use multiple threads to process the remaining mapping pipeline stages. You might want to disable partitioning for a Java transformation when the Java code requires that the transformation be processed with one thread.
You can configure a Java transformation to maintain the row order of the input data by selecting the
Stateless
advanced property for the transformation.
Previously, you cleared the stateless property if the Java transformation needed to be processed with one thread. When the stateless property was cleared, the Data Integration Service did not create partitions for the entire mapping.
Effective in version 10.0, when a mapping contains a transformation that does not support partitioning, the Data Integration Service uses one thread to process the transformation. The service can use multiple threads to process the remaining mapping pipeline stages.
Previously, when a mapping contained a transformation that did not support partitioning, the Data Integration Service did not create partitions for the mapping. The service used one thread to process each mapping pipeline stage.
For more information about partitioned mappings, see the "Partitioned Mappings" chapter in the
Informatica 10.0 Developer Mapping Guide
.


Updated April 09, 2019