It is recommended that you use the following guidelines during performance testing:
Use a repeatable, measurable, and documented process. For more information, see Figure 27.
Perform the following tests:
Response time test
Each test measures different types of performance. Individuals and groups who possess ownership of the architecture and infrastructure typically perform and administer these tests.
To validate the health of the environment, test each custom script you use.
Before you attempt the full user load, obtain a low load baseline. To use this baseline to measure scalability, compare results between the baseline and the target loads.
To prevent excessive resource utilization due to large numbers of users who simultaneously login, start users at a controlled rate. This rate depends on the total configured capacity of the deployment. For every one thousand users of configured capacity, add one user every three seconds. For example, if the product is configured for five thousand users, then add five users every three seconds.
Excessive login rate can cause the application server tier to consume 100% CPU, and logins begin to fail.
To prevent inaccuracies during load testing due to simulated users who simultaneously perform transactions, randomize the think time, which is the amount of time the user requires before the user provides input to the server.
Set randomization ranges according to the relative think times of expert and new users when compared to the average think times in the script.
Figure 27 illustrates the performance testing process.
Figure 27. Performance Testing Process
Developing and Deploying Siebel Business Applications