Databases Reference
In-Depth Information
around the world and no one consolidated database. Some database adminis-
trators don't feel comfortable with a configuration like this and prefer to have
one central consolidated copy to ensure that all database data is secured and
backed up on a regular basis.
Figure 12-7 shows an arrangement of replicated tables based on the princi-
ples mentioned. There are two copies each of Tables A, B, E, and F, and three
copies of Table D. Apparently, Table C is relatively unimportant or infrequently
used, and it is located solely in Los Angeles.
You can take this partitioning of table data down to an even lower level. Most
manufacturers' replication schemes support horizontal and vertical data parti-
tioning. With
horizontal partitioning,
you filter the rows so that a subset of the
available rows is replicated to the remote site. This is used when you want a local
copy of some, but not all, of the rows. With
vertical partitioning,
you filter the
columns so that only selected columns are replicated. This is typically done as a
security measure, giving the remote locations only the data they need to do their
jobs without generating multiple copies of more sensitive data. You can combine
the two types of partitioning and filter a table both horizontally and vertically.
You can see an example of partitioning, which is sometimes referred to as
table
fragmentation,
in Figure 12-8. This figure shows the same network we've
been using, but with Table G added. Table G might be, for example, the com-
pany's employee table: the records of the employees who work in a given city
are stored in that city's computer. G1 is the subset of records of Table G with
the records of the employees who work in Memphis, G2 is the subset of records
for the employees who work in Los Angeles, and so forth. This makes sense
when one considers that most of the query and access activity on a particular
employee's record will take place at his or her work location.
The drawback of partitioning is that when one of the sites, say the New York
headquarters location, occasionally needs to run an application that requires
accessing the employee records of everyone in the company, it must collect the
records from every one of the five sites. One way to minimize this problem would
be to keep a consolidated copy of Table G with all employee records in New York.
FOR EXAMPLE
Distributed Data
As you can see, there is no one best distributed data configuration. The
design, like any database design solution, must be matched to the company,
its data, and the way it does business. There are some general guidelines that
apply in most situations, such as minimizing the exposure of sensitive data
and keeping data communication requirements to a minimum, but even with
strict adherence to these guidelines the available configurations vary widely.
Continued
