Archives of the TeradataForum
Message Posted: Wed, 04 Aug 2004 @ 09:09:23 GMT
My own background with DB2 is next to zero but based on the description of pages you have provided I would think that the Teradata principle of proportional sampling is pretty close to that used by DB2. At least as the mathematical statistics is concerned. Instead of simple sampling a stratified version is used. In case of DB2 it's based on pages; in case of TD it's the AMPs. Proportional samples will most likely not be robust to chi-square (you could probably measure that and compare to what you have on DB2, if you need to) but the I/O is significantly reduced. You are right about AMPs as the entities representing parallelism in TD but in this case they have the second function of natural data stratification.
|Copyright 2016 - All Rights Reserved
|Last Modified: 15 Jun 2023