Archives of the TeradataForum

Message Posted: Fri, 15 Dec 2006 @ 19:28:43 GMT

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

Interesting thought Randall.

Is it possible that some of the speed comes from the AMPs using pure SQL without requiring it to bang around a lot making abstract function calls to verify character data that is suppose to be numeric? uuuumph

Since concatenating and creating artifical primary keys (PK) is what people have always done (bad habits and/or necessity) in other databases, this design and procedures have unfortunately been ported straight to Teradata "as is" with the same messy data and requirement for PK. When people take the time to truly learn Teradata (UPI and NUPI) instead of making Teradata clunky like the other guys, is this throwing up a slippery- slope then need not be there? uuuuumph

Instead, in Teradata we have tended to store numeric data in a numeric data type in an atomic format (each individual element), normalized and verified it at load time using error tables. Then, fix the junky data after extracting it from the error table, fixing it and then loading it properly. It eliminates the need for these verification functions, speeds up data access, eliminates worry of non-numerics in columns that are suppose to be numeric and increases flexibility of data use. Oh yeah, did Randall mention speed? uuumph

Very interesting!

Michael Larkins
Certified Teradata Master
Certified Teradata SQL Instructor

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