Archives of the TeradataForum
Message Posted: Thu, 05 Dec 2002 @ 21:22:02 GMT
Subj: | | Re: Loading Fact data directly onto Teradata |
|
From: | | Dwight Etheridge |
Hi,
You may have hinted at "Teradata Procedures" for ETL. If you are referreing to Teradata Stored Procedures for ETL work - you might not
be thinking in ETL set theory.
If you can turn an ETL processs into a sort/merge problem that the DBC can do in parallel using SQL, you will almost certainly kick
Oracle. For example, lets say a repeating group in a legacy system needs to be stored over multiple rows in a Teradata table. This
transform can be done with a SP using a cursor (not recommended) or using join/case statement to pivot the data (recommended). The latter
workload is almost certainly divided and conquered using all the units of parallelism available. The set approach (latter) will scale
beautifully as the number of rows to process increases.
I recently shut off a massive Informatica/Oracle system after converting the mappings Teradata ETL. Processes that took 10 days now take
15 minutes using parallel techniques and SQL.
|