|
Archives of the TeradataForumMessage Posted: Tue, 18 May 2010 @ 11:51:47 GMT
Dmitriy Boyko wrote:
The main differences: ANSI - default CASESPECIFIC chars/literals - default MULTISET tables - COMMIT Teradata - default NOT CASESPECIFIC chars/strings - default SET tables - BT/ET or implicit transactions Most customers run their system in Teradata mode, but some switch to ANSI for load scripts in BTEQ: ANSI throws more errors, whereas Teradata tries to avoid them, e.g. insert 'abcde' into a char(4) column, ANSI fails, Teradata succeeds, but truncates the string. You'll find all the details in the manuals: SQL Request and Transaction Processing
It doesn't estimate :-) It's the exact number, Teradata creates the full answer set, returns that count and then starts sending row.
If fact tables are mainly accessed through dimension columns you're trying to get "LT/ST" or "star"-joins: SQL Request and Transaction Processing NUPI sizes are a bit hard to pre-calculate, two columns with the same datatype might have totally different sizes depending on rows per val. For column with a large number of rows per val the size will approach 8 bytes per row (10 byte for PPI). You'll find all the sizing details in Database Design Dieter
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Copyright 2016 - All Rights Reserved | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Last Modified: 15 Jun 2023 | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||