Assume that a retail enterprise has a large, nonpartitioned primary index sales table containing the details of each transaction for the previous 24 full months plus the current month-to-date. Once a month, the transactions from the oldest month are deleted from the table. Current transactions are loaded into the table nightly using Teradata Parallel Transporter. Most transactions are added on their occurrence date, but a small percentage of transactions may be reported days later. The number of transactions per month is similar for all months.
Each row contains, among other things, the product code for the item, the transaction date, an identifier for the sales agent, and the quantity sold. The rows are short, and the data blocks are large. The primary index is a composite of product code, transaction date, and the agent identifier.