Home Page for the TeradataForum
 

Archives of the TeradataForum

Message Posted: Wed, 31 Jan 2007 @ 19:42:22 GMT


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


Subj:   Re: How Do You Measure/Validate Compression Savings?
 
From:   John Hall

I feel that something is lacking with this discussion, so let me throw-out a strawman:

Most of the machines that I have worked with had less than a handful of large tables - where a few tables consume the major fraction of current perm. There would be a number of moderate size tables (10's to 100's) and lots (maybe thousands) of small tables (they're small when compared against the large tables).

While compression on the large tables could free significant amounts of space, there would a sharp drop-off in space recovery when trying to compress the moderate sized tables. Compressing the moderate sized tables might still be worthwhile, but it's not going to have the impact that it did with the large tables.

Would be any real value in compressing the small tables? After all, how many small tables do you need to significantly compress to save 10's of gigabytes? Actually, don't we need to save enough space to eliminate one or mode nodes to be worthwhile?

Going through the gyrations of compressing small tables without saving one or mode nodes worth of space seems like a lot of hand waving and a waste of system resources. The idea is not to create keep-busy work for the DBA and Database Designer, but to save the company a significant amount of money by either reducing the size of the machine or making room for additional tables to be added without an upgrade.

Compressing the biggest tables gives the largest return. From my experience, there's usually only a couple of the large tables on a machine - usually they have been some type of history table. The large tables usually have pretty stable data metrics. The compressible columns and data values are usually pretty well known and were done a long time ago (usually at design time).

Of course, things change over time. On the systems that I have been involved with, I would suggest that a review of large table compression might have been necessary only once every year or so.

So my strawman is that if it's the compression of large tables has the greatest value and that there's only a handful of those tables and I only need to do it occasionally, then why do I need a tool when a manual process is probably more than adequate?

So when it comes to justifying the purchase of such a tool, I think that's the question that needs the answer.



     
  <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