Chunking for Parallelizing Large Table Updates

Chunking in Warehouse Builder automates the use of a "divide and conquer" approach to parallelize the processing of large updates. Users enable chunking for a mapping and define chunking criteria to partition the updates. OWB generates PL/SQL code for the mapping, and at execution time, updates are divided according to chunking criteria, a pool of threads is allocated, and the chunks are processed in parallel.

The benefits of applying chunking include:

• Chunking provides the only method of automatically parallelizing PL/SQL code in Warehouse Builder.

• Chunking avoids the need for large rollback segments. Set-based SQL statements for large updates require large rollback segments, because a single set-based statement does not perform intermediate commits.

• Large updates can be performed incrementally, and if interrupted, chunks already processed do not have to be processed again. Without chunking, if a large update terminates for some reason, all processing must be repeated.

  
  

  See Also: The section on chunking data under Using Oracle Source and Target Operators in Oracle Warehouse Builder Data Modeling, ETL, and Data Quality Guide