Has anyone used R3load LOAD "ANY_ORDER" option?
We are testing a migration with a big (32-core) box and this option is listed in Note 1058437 as an option for R3load.
In the note, it asks for us to consult the DB2 documentation, which describes this option:
"Specifies that the load utility can process the input data in any order, resulting in better performance on symmetric multiprocessing (SMP) systems."
"By default, the load utility preserves record order of source data. When load is operating under an SMP environment, synchronization between parallel processing is required to ensure that order is preserved.
In an SMP environment, specifying the anyorder file type modifier instructs the load utility to not preserve the order, which improves efficiency by avoiding the synchronization necessary to preserve that order. However, if the data to be loaded is presorted, anyorder might corrupt the presorted order, and the benefits of presorting are lost for subsequent queries."
http://www.ibm.com/developerworks/data/library/techarticle/dm-0405melnyk/index.html
Since most performance optimizations for R3load recommend exporting the data unsorted anyway, and it has to re-sort on import, this seems like it may be a good option to try out.
I was just wondering if anyone had tried this as I had not seen it any of the optimization studies for DB2/R3load.