Home Page for the TeradataForum
 

Archives of the TeradataForum

Message Posted: Tue, 24 Feb 2004 @ 13:56:30 GMT


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


Subj:   Re: Insert/Select taking too much time
 
From:   Narayan Murthy

The difference b/w the SELECT and the INS/SELECT with the 10million rows, is that there is a redistribution (of the 10million rows) that takes place after the SELECT based on the primary index of the target table - thus 10hrs!

You could try the following if they could help.

1. The elimination of duplicate rows through the use of DISTINCT - which otherwise will be done implicitly since you have a SET table.

2. Since the SELECT takes 10min you could try to EXPORT these rows (again DISTINCT rows) to a file and then IMPORT and INSERT from that file to your target table. By doing so a direct hash distribution of the input rows would take place and this may avoid inter-AMP traffic due to hash redistribution.

3. If you could rewrite the query in such a way that would nullify the redistribution of the rows i.e. to say when you perform the select operation a spool is being built on all amps (or group amps) -now if these rows hash value are the same as the resultant hash of the PRIMARY index of the target table - would eliminate the movement of the rows between AMPs due to redistribution, because the redistributed hash value would be pointing to the same AMP which contains the SPOOL rows - HOPE THAT IS NOT CONFUSING.


Narayan



     
  <Prev Next>   <<First <Prev
Next>
Last>>
 
 
 
 
 
 
 
 
 
  
  Top Home Privacy Feedback  
 
 
Copyright for the TeradataForum (TDATA-L), Manta BlueSky    
Copyright 2016 - All Rights Reserved    
Last Modified: 27 Dec 2016