Database Reference
In-Depth Information
edge. In order to achieve the complete integration requirement, a global data model must preserve
source information for partially integrated databases.
An extended global data model associated with source information requires new set of data ma-
nipulation operations. On one hand, these operations allow us to query the integrated database.
On the other hand, one can make use of these operations to achieve complete database integration.
When complete instance integration has not been performed on multiple databases, it is necessary
to augment source information to the global data model in order to identify where the instances in the
integrated database come from. The source information allows us to:
i. provide the context information to better interpret the non-fully integrated instances;
ii. support meaningful and flexible query formulation on the partially integrated databases; and
iii. perform entity identification and attribute value conflict resolution within queries or applications
if the need arises.
A number of different data models have been proposed for multidatabase systems (MDBSs). They
can be broadly classified into three main categories according to the degrees of integration:
Type-1: These MDBSs choose not to handle any semantic heterogeneity, e.g. MSQL (Litwin,
Abdellatif, Zeroual and Nicolas, 1989; Lakshman, Saderi and Subramanian, 1996; Wang,
Madnick, Wang and Madnick 1990). In other words, they do not provide global integrated sche-
mas over the preexisting databases.
Type-2: These MDBSs may support global integrated schemas but not integrated instances. In
these MDBSs, the pre-existing database instances representing the same real world objects are
not entirely integrated together (Agrawal, Keller, Wiederhold and Saraswat, 1995; Liu, Pu and
Lee, 1996)
Type-3: These are MDBSs that integrated both the pre-existing database schemas and instances
(Clements, Ganesh, Hwang, Lim, Mediratta, Srivastava, Stenoein, Myriad, and Yang, 1994).
In (Agrawal, Keller, Wiederhold and Saraswat, 1995), a multidatabase is defined to be a set of flex-
ible relations in which local instances that represent the same real-world entities are stored together
as group of tuples. Hence, some implicit grouping of tuples in a flexible relation is required. Flexible
relations also capture the source, consistency and selection information of their tuples. A corresponding
set of flexible relational operations has been developed to manipulate the flexible relations. Neverthe-
less, flexible relational model is not a natural extension of the relational model. Furthermore, the join
between flexible relations has not been defined. A universal relational approach to model and query
multidatabases is proposed in (Zhao, Segev and Chatterjee, 1995). In this approach, a multidatabase
is a universal relation instead of a set of relations. Queries on the universal relation are translated into
multiple local queries against the local relations. The final query results are formed by unioning the
local query results. Source information is attached to tuples in the final query results to indicate where
the tuples come from. However, the source attribute is included in neither the universal relation nor its
query specification. Joins and other operations that involve multiple component databases are not al-
lowed in this model.
Search WWH ::




Custom Search