Archives of the TeradataForum

Message Posted: Thu, 30 Oct 2003 @ 15:11:59 GMT

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

In general, string operations tend to be CPU intensive but the statement's total run time will depend on how the operations are used. For example, if join a large fact table with a few small dimension tables and use concatenation, substring etc on the columns that are not used in the join, the overall string operation overhead might be insignificant. If such operations are applied to the columns before the join and then the jpin is made on them, then that might make the optimizer change the execution plan and go for, say, full table scan instead of SI. That would likely lead to the significant increase in both CPU usage and disk I/O.

BTW, I am not familiar with Larry's theory. If I understand the ratio's definition correctly, two queries with

CPUtime1 = 10*CPUtime2 and diskIO1 = 10*diskIO2

seem to be equally efficient in terms of this ratio. Is that right?

Also, when you ask "Does the string operations or case statement cause to use more CPU?" what exactly do you want to compare? Like what would be an alternative to, say, substring? Some intermediate temp tables? Please elaborate.

Regards,

Victor

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