Coding Performance into the CML

The ways an application is designed and implemented affect both the way it performs and the degree to which it scales. This section covers both issues — performance and scalability — in the Corporate Media Library (CML) application.

The Fitch & Mather (F & M) Corporation has 4,500 employees, which means 4,500 potential users of the CML application. Internal research showed that only 1 percent of them are likely to search for books online simultaneously. The developers of the CML knew that 45 users, though very small compared to possible usage rates for an Internet Web application, was probably reasonable for a company the size of F & M.

Still, they designed the CML with performance and scalability in mind, in case usage spikes occur. To this end, they used the following generally recommended model:

• Referential integrity rules are enforced neither in the scripting code of the business logic nor in the Visual Basic® code of the data access components, but in the database itself. See Referential Integrity in the CML.
• The client computers' resources are used whenever possible to reduce the Web server's workload. See Leveraging Client Resources.
• All data is stored in the lowest (data-services) tier. See Storing Data on the Data-Services Tier.
• The business logic executes in a separate tier from data storage. See Executing Business Logic on the Middle Tier.
• The business logic executes faster because it is implemented as Component Object Model (COM) components, not as script. This is faster because COM components are compiled, not interpreted.
• Stored procedures, which reside in the CML's database, are used wherever they make sense. See Stored Procedures in the CML.

Even with the best design, every application must function within the limits of its design and environment. The CML's limits are as follows:

• Theoretical limits of the CML. Because the CML is an intranet application tailored to 45 users, it cannot and need not scale to large proportions. There is nothing in the CML's code, however, that sets limits, for example, as in the number of simultaneous users, user requests, or size of the database.
• Practical limits of the CML. The most probable bottleneck affecting scalability or performance is remote data service (RDS) traffic, specifically at the point where RDS communicates with ActiveX® Data Objects (ADO).