Data Communications Requirements of Distributed Database Systems - High-Performance Web Databases: Design, Development, and Deployment - page 467

a centralized DBMS. Performance transparency depends on a distributed
query optimizer — which finds a heuristically optimized plan to execute
the command — and the network equipment over which data is trans-
ferred. This issue is also closely related to the determination of the data
communications requirements of distributed DBMSs; three approaches to
this issue are presented further in this chapter.
PROS AND CONS OF DISTRIBUTED DBMSS
Distributed DBMSs have several advantages over conventional DBMSs,
including the following:
• A distributed DBMS reflects the changing structure of the organiza-
tion. Organizations today are characterized by a decentralized struc-
ture spread across many geographic locations. The modular
implementation of distributed databases allows data to change as the
organization changes. Incremental growth and upgrades are made eas-
ier. Whereas changing the structure of the data in a centralized system
necessitates taking down the system, adding new nodes to a distribut-
ed system has minimal impact.
• Because a majority of processing is done at the local level, faster re-
sponse time can be achieved. Also, by locating data near its users,
communications overhead can be reduced. In a centralized database
environment, communications overhead is the key determinant of re-
sponse time.
• In a distributed environment, greater reliability and availability can be
achieved. In a centralized environment, failure at the central site
would cause total system failure. If data is distributed, processing can
continue after a failure, saving valuable time and money.
• In full replication mode, a Distributed Database System supports di-
saster recovery by locating data at physically different sites.
• With the workload distributed to different sites, the system workload
becomes better balanced. Shifting the workload to a smaller system
makes economic sense because computing on a mainframe can cost
more than $100,000 per million instructions per second (MIPS).
Despite the numerous advantages of distributed database system, there
are several problems inherent to having data located at multiple sites.
Most of these problems stem from transparency issues discussed earlier.
Among them are the following:
• Because data is no longer in a centralized environment, security be-
comes a greater issue. Security must be addressed consistently at
each site.
• The high cost of communications equipment, dedicated lines, or net-
work access needed to move a large amount of data between multiple
sites.