Data Access and Transactions
***** *Distribution and Scaling Issues
Hosting a large volume of business transactions from clients around the world introduces a demanding set of programming and administrative requirements. Distributed applications very often rely on diverse resources beyond the applications scope of control. Web applications must be prepared to host a large volume of connections without suffering a significant loss in performance.
The key factors that influence a components ability to service a large volume of users include:
Resources allocated per user (memory, data connections).
Amount of information retrieved from and sent to the browser.
Location of processing (client, server, or both).
Impact of component distribution topology on response time.
Time to process common requests.
Components should be located as close as possible to the data source. If you are building a distributed application with a number of COM component packages running on local and remote servers, try to group your components according to the location of your data. For example, the Accounting server should host both the Accounting Component Services package and the Accounting database.
Pool your resources by server process. Note that Component Services runs each hosted package in a separate server process. The fewer pools that are running on a server, the more efficiently resources are pooled. Try to group components so that they share expensive resources, such as connections to a specific database. If you reuse these resources within your package, you will improve the performance and scaling of your application. For example, if a database lookup and a database update component are running in a customer maintenance application, package those components together so that they can share database connections.