Home Page for the TeradataForum
 

Archives of the TeradataForum

Message Posted: Tue, 10 Apr 2007 @ 12:59:34 GMT


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


Subj:   Re: Merging large volumes of data into large history tables
 
From:   bob hahn

Insert/Select also works at the block level--the recent changes mentioned also handle USI/NUSI maintenance (and rollback) at the block level.

Fastload only populates empty tables which is much faster. Even here, Insert/Select to an empty table is faster even than just the Fastload apply phase excluding acquisition.

For both Multiload and i/s, a key factor is the block hit rate.

Some recent tests Ive seen comparing i/s with Multiload application phase assuming 10 hits per block measured:

16,752 rows/second/node for i/s, 7901 for Multiload.

12,169 rows/second/node for i/s with a NUSI (100 rows/value), 5162 for Multiload.


Another set on prior hardware generation using 5 hits/block:

6,378 rows/second/node for i/s, 5306 for Multiload.

5,222 rows/second/node for i/s with a NUSI (100 rows/value), 2600 for Multiload.


Insert/select also avoids the client/database resource consumption of FastExport (spool + 2 redistributions) and Multiload acquisition phase.

These were standalone tests with 100 byte rows and presumably other differences from your case.



     
  <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: 15 Jun 2023