Host-level Architecture for GWD

The application architecture for GWD should be designed with consideration of the following points:

• Modularize the application so that each module can be assigned to different hosts based upon its scalability needs
• Partition the application across hosts by considering the communication among modules on different hosts and utilization of data among these hosts
• Consider how best to balance the computation and communication as well as dealing with the possibility of congestion

These are important points to address in order to realize high-performance and scalability of GWD. The overall application architecture of GWD is as follows:

GWD offers a variety of features which can be split up into modules primarily determined on the basis of different features. Each modue can be further divided into smaller sub-modules and then can be assigned to different hosts as needed. Modules by features can also be grouped by their functions as shown in the diagram (see Modularization for partitioning.)

In light of the analysis on concurrency, scalability and congestion for GWD, now let's look at one possible implementation of the host-level architecture focusing on the site management feature.

To mitigate possible congestion and realize scalability, the module can be divided into smaller sub-modules. Considering the communication among these sub-modules and databases, they are assigned to different hosts as follows:

Middleware can also contribute to better performance of the application. For example, MOM (Message Oriented Middleware) offers a solution to combat possible congestion.

Use of mobile code can similarly enhance the application's scalability because the server's load is decreased by moving some application logic to the clients. Especially because GWD uses a web browser as the primary interface, mobile code can easily be implemented by embedding Java applets in the HTML representation on browsers.