Archives of the TeradataForum
Message Posted: Thu, 20 Jan 2005 @ 14:09:12 GMT
My information comes from an incident that I worked on a few years ago (on V2R4). I accept that things do change over time and this may be one of those but I don't believe so. What happens is this:
- the optimiser looks at the SQL, DDL etc and determines if fast-path is ALLOWED. Note that this is ALLOWED, not that it will necessarily happen.
- when each AMP starts the Merge step it checks if the optimiser ALLOWs fast-path.
If it is not allowed then each row is transient-journalled.
If it is allowed then each AMP checks the contents of the target table on that AMP (I think this is done using the MI and CI rather than the table header)
If the target table is empty on a particular AMP then that AMP will use fast-path
If the target table is not empty on a particular AMP then that AMP will not use fast-path
Yes this means that some AMPs will use fast-path and some will not. I have never seen this documented in any manual
|Copyright 2016 - All Rights Reserved|
|Last Modified: 28 Jun 2020|