Best way to generate the Sequential numbers : CSUM or IDENTITY columns? - response (8) by dnoeth

Hi Dheem,
when you check the QueryLog after running a CSUM(1,1) you'll notice that a single AMP (usually vproc 0) processed all the data, high cpu/io and spool.
For OLAP functions there's a distribution f the data based on PARTTITION and ORDER and for CSUM(1,1) there's only 1 partition and no order.
 
Hi Mike,
when you are running out of numbers it's not a limitation of Teradata's IDENTITY, you just used the wrong datatype.
BIGINT should be enough, otherwise there's DEC(38,0).
And the limitations on load utilities have been removed since a few releases.
Dieter