Archives of the TeradataForum
Message Posted: Thu, 05 Dec 2002 @ 22:41:16 GMT
Subj: | | Re: Loading Fact data directly onto Teradata |
|
From: | | Terry Stover |
You can trap the data type errors in the fast load, the bad records go to an error file. At my old job we had a perl "fixer" that
corrected the common errors and reloaded the corrected records and the error stats. Fast load can check constraints, I assume it would
check RI constraints so you could pick up new keys with a 2 pass process.
WRT delete / insert to bypass transient journal, if you have the space you can do a new table then a rename when it's loaded. With
continous availability you'd probably have to boot everyone off to replace the production table, the access lock for read will block any
ddl.
I've used Informatica but not against Teradata. You might be better off outputting Informatica to a text file then doing a fast load.
With sqlserver we found bcp to be 2-3x faster than Informatica loading the db directly.
Finally, NCR advertises TPUMP for the continuous availability / continuous update problem. I don't know anyone who has used it
though.
|