Managing Multiple Databases Across Heterogeneous Hardware and Software Systems - High-Performance Web Databases: Design, Development, and Deployment - page 480

With multiple databases, one data-
base must be considered the master database. That is, the values of the
data items in the master database are considered to be the final word. This
designation as master, however, is on a data basis. For example, the data-
base on the corporate mainframe may be considered the master for
customer name and address, whereas the shipping database (which also
contains the customer name and address) is considered the master for the
shipping date for a customer's order.

Designation of the Database Master.

In the determination of the master database, the following criteria
should be considered:

• Stability. How stable is the database? How much maintenance is re-

quired?

• Vendor support. How effective is the vendor support for this sys-

tem/database? How promptly does the vendor respond to support
calls?

• In-house expertise. Who within the organization knows the system/da-

tabase well enough to answer routine questions and solve perfor-
mance problems?

• Available consultant expertise. If in-house expertise does not exist or

should disappear, is there a pool of consultant expertise upon which
to draw?

• Volume of data. What is the current volume of data on the database? If

data fields are added, what are the ramifications?

• User load on the host system. How will the transactions affect system

performance? Batch updates, for example, can almost negate online
response while it is running. Again, an estimate of transaction volume
should be made.

Making Critical Design Decisions

The design phase of the project should not be
entirely end-user driven. The end-user maintains a biased view of the data
and often requires data items that are actually derived information and not
necessarily stored as fields (e.g., average customer order size). A database
administrator (DBA) view, in this case, is required.

Overall Data Architecture.

The DBA should obtain information about the data items in question
from the end-users because they know the data. However, the DBA should
then take that information and put it into the context of the overall data-
base structure.

For example, the users tell the DBA about a data item called
Last_Contact. The DBA is required to find out the context of this field (i.e.,
contact by whom?). This may be different from the data item in the main
database with a similar name.