Getting ready

Let's connect to our local database instance and create a table using SQLA. Remember, PI is the only index that is not only used for data distribution but for retrieval also. Choosing a PI is the most important task. Bad PI has a ripple effect on queries. The following are some pointers for choosing a better PI:

• Consider data distribution
• The more unique the PI, the better the distribution across all AMPs on Teradata
• Understand how will the table be joined to
• Using the most frequent join path will improve joins to the table
• However, this column may be highly skewed