Partition method for creating key change column

[udayanguha]

Hi,
I am trying to create a key change column through sort stage. In the partition tab, shall I specify it as auto partition and Datastage will take care of the best partitioning method or shall I explicitly mention a hash partition in the property? I have heard different views from people. Some people suggest to always mention explicitly the partition method and some suggest to leave it as auto. A bit confused now what to use?

[ray.wurlod]

Auto will give you Hash on the (entire) Sort key. It may be more efficient to specify explicitly under a couple of circumstances.

• If there is high cardinality on the first Sort key, you may prefer to partition on that key only.
  
  If the Sort key is an integer, then the Modulus algorithm will be more efficient than Hash.

[ray.wurlod]

Datastage will take care of the best partitioning method

Not quite true. DataStage will select a partitioning method that will always work. It may not be "best". It will be guaranteed to partition the data correctly for the stage in question.