Home Page for the TeradataForum
 

Archives of the TeradataForum

Message Posted: Mon, 11 Nov 2002 @ 06:14:37 GMT


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


Subj:   Re: 3NF vs Dimensional modelling
 
From:   Robert Doss

Hi Neil,

I don't really disagree with you, but here are some comments on your comments.


  In the case of ROLAP's like Business Objects or Microstrategy, all of this SQL complexity is masked from the requestor. I don't care if it uses 20 temp files and does an outer join, I only care that the visual object model I use to develop the request is powerful enough to let me ask any question I want, and then let the SQL generator do its job.  


OK. But wouldn't you agree that the need for these (expensive) tools to do this in the first place indicates a failure of relational technology & SQL to address analytical requirements? As you state:


  Now, the time it takes to satisfy the query is another issue, but that goes to the heart of the problem of why RDBMS's are so poor at analytical processing.  


  MOLAPs are one answer, but we all know now that they can't scale.  


I see this changing. I have worked on two installations where we successfully scaled up Analysis Server to handle very large databases. I am aware of quite a few other big sites doing this as well. I won't say it was easy, but then scaling Teradata or DB2 beyond a few gigabytes circa model 2 and 3 and DB2 v2.3 wasn't trivial either. The point is that MOLAP technology is maturing at about the same pace that relational did. Microsoft AS is big step forward; I expect others to follow shortly.


  you better learn to write your queries in MDX, not SQL. Maybe that's the long term answer, I'm not sure.  


I have, and once I got over the fact that MDX is not SQL I find it a much more natural way to express analytical queries. Instead of an expensive tool that makes SQL and realational technology do unnatural acts, you can use MDX to do exactly what you want directly against the data. In my opinion, MDX (or something like it) is part of the answer. BTW - it is pretty simple to hide the MDX from users so they never even know it exists.


  But in the meantime, I'd take issue with your comment that we don't need more OLAP operators in SQL, I think we do. And we do especially because that forces the database vendors to build optimizer technology to deal with them.  


Sure, I agree, I welcome any power I can get. However, after using MDX these functions seem very un-natural, not very flexible or performant and overly complex to me. All I was trying to express is that I believe there may be better long term solutions / architectures for doing analytical processing.


Thanks for the feedback,

Bob



     
  <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