Archives of the TeradataForum
Message Posted: Fri, 22 Feb 2002 @ 21:48:03 GMT
| Subj: || || Re: NUPI considerations |
| From: || || James A. Judge |
What about a NUPI of (col1,col2)? Assuming you will be using the calendar table in most queries with the 1.5M row, and using a narrow
range of (dates and/or rows in the 1.5M table). Then a Cartesian join between the calendar table and the 1.5M to get all selected col1,col2
combinations would be doable to get a redistributed spool of (col1,col2) for local amp joins to the 1.5B row table (but that spool size
grows exponentially with the increase in selected dates and 1.5M table rows).
Although the 1000 synonyms (if really that even and not a few 100,000+ values) wouldn't bother me too much when joined to a UPI value of
col2. I think elapsed time added for additional USI/NUSI for table maintenance would be more than the synonym overhead.