Overview

One of the major differences between a classical OLTP systemand a data warehouse is the maintenance process.A person using on-line screens normally maintains thedata in an OLTP system. When these users create orchange the data, they demand a very quick response timefor each transaction. There may be several hundred people each creatingan individual OLTP transaction that must be processed at any onepoint in time. There will be peak processing times that the system mustbe tuned to handle. By contrast, load processes maintain a data warehouse.All the OLTP transactions that were processed during the dayare essentially reprocessed. The data warehouse must process the veryhigh volume of data that arrives at the same time, usually at night.Those same couple of hundred OLTP transactions will turn into tens ofmillions of transactions that need to be applied to the data warehousedatabase. This means that the load processes will be a very importantpart of a data warehouse. This chapter will discuss these loadingprocesses, but more importantly will offer some tips that you may beable to use to reduce the development time of your first data warehouse.Finally, I am going to discuss a more advanced technique thatcan be used to smooth the database maintenance workload over a dayfor continuous 24-hour data maintenance.

Out-of-the-ordinary processing does not mean that the quality of the data in the database should be sacrificed.

I am going to discuss some techniques you can use to speed up themaintenance of your data warehouse implementation. However, all ofthese techniques will not work...