ABSTRACT
The invention of data warehouses was inevitable in order to reduce conflict between operational Online Transactional Processing (OLTP) databases and the large disk access requirements of reporting. OLTP databases need heavy resource sharing among lots of users, and data warehouses need to use those same resources for extended periods of time, which is why data warehouses exist, because the two functions can cause debilitating conflict among CPU, RAM, and I/O resources. OLTP databases can be altered and manipulated much more easily, as they are designed for a constant stream of small adjustments from a very large group of users. Surrogate keys are replacement key values, which usually make database access more efficient. A standard normalized relational database model can be completely unsuitable for a data warehouse, and sometimes even a denormalized relational database model might not make sense.
