DISTRIBUTED DATABASE SYSTEMS - Database Design and Development: An Essential Guide for IT Professionals - page 557

Partially connected. Only certain sites are linked. Sites with high volumes of data
traffic are usually connected. Less expensive than fully connected configuration.

Bus. All sites connected to a backbone link. Simple, easily extendable, and cost-
effective. Useful in small network with low traffic volumes.

Star. Used for connecting sites with the use of the concept of a central site. The
configuration fails if the central site goes down.

Ring. Connects sites in a closed loop. Generally used for high-performance

networks.

Data Distribution

All along we have been saying that parts of the database are stored and managed
at the various sites of an organization. We looked at the levels of schemas that must
be present at the sites for representing the database. We also discussed the network
configurations that could enable moving of data between sites as needed. Now the
question remains, How exactly do you break up the database into portions that can
be kept at different sites? How do you divide the data content and distribute it
among the sites?
Two basic methods are commonly used to break up the database for the purpose
of distribution among the sites. Our discussions will be based on the relational data
model. Nevertheless, the principle applies equally well to the other data models such
as hierarchical or network. Let us consider just one relation, an EMPLOYEE rela-
tion, to present the methods.
Figure 18-9 illustrates the following two methods using the EMPLOYEE rela-
tion as the contents of the database:

• Data fragmentation
• Data replication

Data Fragmentation When using the data fragmentation or data partitioning
method, you break up a relation into smaller fragments or partitions in one of two
ways. Look at the data in the EMPLOYEE relation. What is the data content made
of? Rows and columns. You know the significance of rows and columns in a rela-
tion. How can you partition the relation into fragments? You can break up the rela-
tion horizontally into groups of rows or vertically into groups of columns. You can
then store these fragments at the various sites.

Horizontal fragmentation Each fragment consists of a subset of rows or tuples of
a relation. How can you use this method to create fragments from the EMPLOYEE
relation? Let us say that three sites exist on your distributed system network—New
York, London, and Milan. Naturally, the Milan site would be interested in dealing
with the Italian employees, the London site with the British employees, and the
New York site with the American employees. So it makes sense to break up the