Archives of the TeradataForum
Message Posted: Tue, 09 Jul 2002 @ 15:14:43 GMT
| Subj: || || Re: Uniform Distributions from RANDOM number function |
| From: || || John Hall |
I agree with Jose. Given what you're trying to accomplish, it's better to use an Inmod to generate a sequential number (both the random
number or hash approach will give you repeats).
I would also argue that the sequential number approach is going to be your 'safest' approach. It's most likely to work the first time
and is probably the easiest to deal with over the years.
I stumbled a bit over your statement about "generating a large volume of sequential numbers in parallel". Are we talking about a single
job running at a time (like the fastload) or multiple jobs running concurrently? If we're talking about the fastload, then I don't see
where you're generating unique numbers in parallel. Although fastload is a parallel process, the incoming rows are read one at a time and
that's where your Inmod comes into play.