Case Study 4: Unique Primary Index Values
If a table has many more unique values than the system has AMPS, a UPI always provides an even distribution of rows across the AMPs. To ensure efficient space utilization, you should define the primary index for a table on a unique column unless you plan to hash the rows of a minor entity table to the same AMPs as matching rows from a major entity table in order to enhance join processing. Be aware that join indexes might be a better solution, depending on the particular application. See Chapter 10: “Join and Hash Indexes,” for more information.
The illustration shows an ideal distribution of customer table rows when a UPI is defined on the system-assigned primary key, the customer_number attribute.