Databases Reference
In-Depth Information
A special case of client server processing occurs when the client runs on the same
machine as the server. This is known as “local client” processing.
In general, the client server model is known as a two-tier architecture, where the cli-
ent resides on the first tier, and the server is the second heavier tier. Two-tier models
tend to scale well up to roughly 100 clients. Beyond that contention issues typically
require expansion of the server to a shared-nothing architecture of some sort, or the
introduction of a third tier.
Figure 13.4
Three-tier database architecture. (Image courtesy of IBM.)
To resolve the limitations of two-tier architectures the industry began to adopt a
number of three-tier architectures, as shown in Figure 13.4. These architectures separate
the workload balancing and transaction flow from the database layer. The client tier and
the database tier remain, but between them a new tier is inserted. The middle tier, often
called the application server, includes the bulk of the application processing logic as well
as transaction routing and queuing technology. The transaction handling usually also
supports an asynchronous mode.
One of the key values of the application server concept is that it lends itself to load
balancing, so that the middle tier may contain several physical servers that each assumes
part of the load of the middle tier. The application server does not drive the graphical
user interface (GUI) at the client, and similarly because the bulk of the business logic is
executed at the application server, the client tier becomes predominantly a GUI with a
