Home Page for the TeradataForum
 

Archives of the TeradataForum

Message Posted: Tue, 03 Feb 2004 @ 18:07:05 GMT


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


Subj:   Re: Query Performance
 
From:   Sirek, Mark W

There's not much you could have done to avoid the skewing. The optimizer currently assumes an even distribution of row values among the AMPs. But when the 444 values for branch_id are hashed to create spool 6, there will likely be some AMPs that receive many more rows than other AMPs because the number of values is probably not that much greater than (within an order of magnitude) the number of AMPs on your system (or perhaps even less). The sample size (# of values) needs to be a lot higher than the number of buckets (AMPs) in order for an even distribution to occur.

But a bigger problem seems to be that spool 6 is estimated in the explain to be 996 rows. This is apparently far from the actual value. If all relevant statistics have been collected, perhaps it's time to see if a more recent database version has a fix that addresses this estimation problem.


- Mark



     
  <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