Primary key vs non unique index performance

Due to a programming error during a bulk load, there are certain tables in the
database I'm working on which have duplicate entries loaded. When the bulk load
ends, duplicates sometimes exist in the table and the primary key index refuses to
rebuild making the table unusable.

  A management decision has been made to replace the primary key index with a non
unique index that uses the same columns. According to the person who made this
decision, the cost in performance between a select using the primary key and a select
using the non-unique index will be negligible. The number of duplicates inserted in the
table is actually quite small and according to management the data error caused by
the duplicates is acceptable. But, I have trouble justifying the existence of a table that
does not have a primary key. Typical selects on this table will return less then 5% of
the tables data.

  Does anyone have statistics or reports that demonstrate the difference in
performance between a primary key index and a non unique index? Is this a cause
for serious concern for database performance? Received on Fri Jan 26 2001 - 22:13:37 CST