A Model and Language for Bitemporal Schema Versioning in Data Warehouses

Abstract

A Data Warehouse (DW) is a vast collection of historical data built to support multidimensional data analysis applications. In this context, an important problem is that of evolving the implementation (multidimensional, relational) schema of a DW to incorporate new requirements. This paper introduces a conceptual evolution model based on bitemporal versioning of multidimensional schemas, which allows one to modify the DW schema (a) in an implementation-independent manner, and (b) without affecting the operation of existing applications.It also presents a SQL-like language associated to this model, which offers expressions to create and change versions of multidimensional schemas.

Temporal and Evolving Data Warehouse Design

“…The prototype is based on the concept of augmented schema to increase crossversion querying flexibility. Another observation is that two studies, [5,21], extend SQL for querying multiple versions of a DW. Rechy-Ramírez and Benítez-Guerrero [21] propose an SQL-like language that allows users to express evolution requirements.…”

“…The augmented schema is stored together with schema versions. Other than augmented schema, some approaches, for example, [21], propose an extension to SQL for querying a DW. Once data has been retrieved, the interpretation of results is also challenging.…”

“…Among these schemes, in [21] a conceptual model is proposed for DW schema that facilitates the modification of a DW schema in an implementation independent manner, without affecting its operations. Similarly, authors in [29] handle slowly changing dimensions as well as state-oriented data without fundamentally changing the design.…”

“…(ii) For handling changes in a DW (recall schema and data changes), a number of efforts [17][18][19][20][21] have been made. According to Eder and Wiggisser [27] and Eder and Wiggisser [28], there are two ways to handle these changes: (a) updating schema, transforming existing data into the new schema, discarding the schema, and using the updated schema for future data population: this is called evolution scheme [2]; (b) creating a new schema, maintaining both schemas, and using only the new schema for future data population [24]: however, both schemas can be used for retrieval from the DW.…”

“…To summarize, the existence of TDBs and changes in them lead to the development of a Temporal Data Warehouse (TDW). A TDW requires consideration of several aspects in its DW, including temporal support in its data models [12][13][14][15][16], changes to its schema, and changes to its data [17][18][19][20][21].…”