Archives of the TeradataForum

Message Posted: Wed, 18 Jun 2014 @ 22:14:52 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

Depending on your underlying data, it's easy to imagine that query returning a high percentage of rows given all those ORs. That might make a product join a reasonable choice.

If you re-write it using and/not, do you get better performance? Something like this, assuming I have my or/and/equals/not equals/not logic right:

     (
      VWDB.TAB1.ID_TYPE_CD  = 1
      AND not
      ( -- rows where all these are the case are the only ones to exclude.
       VWDB.View1.PRIM_ID_IND  <>  'N'
       and
      VWDB_View1.STAT_CD  =  1
       and
     VWDB.View2.PRD_STATUS_CD  =  6
       and
      VWDB.View2.DW_VER_END_DT  =  {d '9999-12-31'}
       and
       VWDB_VWS.View2.ACCT_TYPE_CD  =  250
       and
     VWDB.View2.TAX_PRD_END_DT  =  {d '9999-12-31'}
      )
     )

Then maybe collect stats on (PRIM_ID_IND, STAT_CD ) and (PRD_STATUS_CD, DW_VER_END_DT , ACCT_TYPE_CD, TAX_PRD_END_DT) and see if that helps.

Dave

	<Prev	Next>		<<First	<Prev	Next>	Last>>