Archives of the TeradataForum
Message Posted: Fri, 12 Mar 2010 @ 15:11:30 GMT
You should have gotten all the software with your Teradata CDs.
Ask your local DBA or Admin to provide you with the software, or if you have access to the Teradata site (Teradata at your service, not the public one) you can download the requisite patch versions there.
I would definitely consult the manuals on TPUMP, MLOAD, and Fastload before forming an ETL strategy, they are all different tools suited for different purposes.
As for performance:
I am assuming since you are using BTEQ your source data is in a file, so my response takes that into account.
The upper bounds for TPUMP used in Mini batch , for an average sized (wide) table that I have been able to achieve is about 4000- 5000 rows per second, when the source is from a file. Tpump is a bit tricky to set up and use, as there are several tuning parameters, and performance is very data dependent.
RTM carefully and test multiple turntables to find the "sweet spot" for performance. It has easier start, restart, capabilities as well.
If anyone has gotten much better performance than that please share your tips.
Fastload runs much differently, can only load empty tables, cannot load through views, and has a litany of other restrictions, and general lack of functionality. It does not run in the SQL partition, loads data via blocks in parallel sessions rather than Multistatement requests as does TPUMP.
But it is screaming fast, and that is what it is meant to be.
The sky is really the limit on rows per second. It depends on the source systems or client throughput, size and business of the Teradata system, networking etc...
But 10s of Thousands of rows per second should be easily achievable if not 100s of thousands.
Eric W. Barner
|Copyright 2016 - All Rights Reserved
|Last Modified: 15 Jun 2023