Short answers:
#1: Yes, No
#2: No
The partitioning on dates fits your example queries, they will benefit from the PPI, but when the PI is accessed it will be slower than before: it's still an single-AMP access, but each populated partition has to be probed. When th PI is heavily used you might consider an additional USI/NUSI on EMPLOYEE_ID.
The distribution is stil based on the PI, just the internal order changes from ROWID to partition/ROWID.
TRUNCATE doesn't exist in Teradata, but a DELETE without WHERE condition is quite similar (so-called FastPath Delete).
Just be shure to collect stats after the load.
Dieter
↧