The process of removing a precise set of infrequently used reference data from an overloaded relational database and keeping it active in an archive where it can be easily and quickly retrieved when needed.
For example, a credit card company acquires large amounts of mission-critical data on its customers - names, addresses, balances, account numbers, interest rates. This information is critical to the company's daily operation. When a monthly bill is issued, that data can be put into an active archive because the company does not need that information unless there are questions about the bill. So the data is taken out of the relational database where it would slow the system down but left available so that it can be quickly accessed when needed.