Archives of the TeradataForum
Message Posted: Mon, 05 Oct 2015 @ 15:44:20 GMT
| Subj: || || Re: Dropping Primary Index |
| From: || || Walter, Todd |
What are you trying to accomplish?
In the original post it sounds as if there is an impression that adding the PI after the data is loaded would improve load performance or
perhaps that the index build would be cheaper if done after the load was finished. If something like this is the impression and therefore the
goal is to build the index after the load - then the impression is incorrect. The PI is a fundamental structure of the table It cannot be built
independently from loading the data into the table. It defines how the data is to be loaded into the table including both how the data is
distributed to the AMPs and how the data is ordered in the table. Having the UPI, NUPI or NOPI defined on the table is fundamental to how the data
is loaded. Because of this, it is not possible to change/add a PI because that would require completely reorganizing the data in the table.
Instead, to change the organization of the data requires that a new table be defined with the desired PI and then perform an INSERT SELECT to put
the data in the table with a very different organization.