Tuple source relational model: A source-aware data model for multidatabases

In some integration projects, complete integration of database instances may not be necessary. It may also be too costly and impossible to do so due to poor local data quality and insufficient instance-level knowledge. In this research, we study how multidatabases with global schemas should be represented and manipulated when the data instances from the local databases do not require to be fully integrated. We propose the tuple source (TS) relational model to represent multidatabases under such an integration requirement. This model extends the classical relational model by augmenting every relation with a source attribute to identify the local database that the tuples come from. The source attribute can also be used to specify the right context to interpret global data instances. To manipulate TS relations, we have developed a set of tuple source relational algebraic operations and an extended SQL query language known as TS-SQL. With TS relational model, flexible multidatabase queries that involve instances from different local databases can be formulated easily. In this paper, we also reported our distributed query processing and optimization strategies and their implementation. (C) 1999 Elsevier Science B.V. All rights reserved.