Home Page for the TeradataForum
 

Archives of the TeradataForum

Message Posted: Tue, 09 Jul 2002 @ 14:44:25 GMT


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


Subj:   Re: Uniform Distributions from RANDOM number function
 
From:   Jose Lora

If you really want a unique number and you're planning to use Fastload, please, forget about the random function that wont give you a unique number (according to the definition of the random function) and use an INMOD routine.

I don't know about you workload, but I use to load a 2 million records each day using Fastload with an inmod routine to generate the sequential numbers and I didn't notice any performance hit. You can use several strategies to get the current maximum id (bteq, C/SQL, perl, etc.). We kept the maximum number on a table to save us some time.

Of course, I'm assuming you only have one batch source for your table and you're planning to Fastload and then Insert/Select the data. If that's not the case, you can not use this approach without serializing your jobs.

By the way, some of the ETL tools can do the same job, for example with SAS/Fastload, you can define the file as a SAS dataset, define the extra calculated field and load the file using the SAS version of Fastload (a really good alternative to NCR Fastload). With Genio from Hummingbird, you can define the input file, make all the transformations that you want (row level) and then stream the data directly to a Fastload (Genio uses an INMOD interface to Fastload and the other Teradata utilities).

The three alternatives are good enough, because they use only CPU and memory time (avoiding the expensive I/O processing that use to take most of the ETL time ).

Jose Lora
Systems Architect
Meredith Corporation



     
  <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