Concurrency in the Data Warehouse

When a data warehouse is loaded at night and queried during the day, there is no requirement for concurrent update and querying. However there are a number of situations where concurrency is needed: trickle feed applications, correcting exception data from the nightly load, the narrowing load window. The end point of the narrowing load window is a data warehouse that is available 7x24. Query Priority Concurrency is the concurrency mechanism implemented by the Informix Red Brick Decision Server. It is called Query Priority Concurrency because it uses versioning to achieve the goal that query performance is unaffected by concurrent loads. The paper discusses the differing requirements for concurrency in a data warehouse, explains why versioning is appropriate, gives a sketch of the implementation and discusses the 6 lock modes that are needed to achieve concurrency and serialised execution. Finally, the frozen query feature is described. This allows users to query the current published version of the data warehouse while the administrators go through all the steps of loading and verifying new data to create the next issue of the warehouse for publication.