When you run a schedule in SoftLayer, the location of the data center and the type of host used for agent virtual machines affects the measurement of response time. For example, if you run the schedule on agent virtual machines that use shared hypervisors, other guest operating systems would also be consuming hardware resources. This action might affect the measurement of response time.
Similarly, in a VMware setup, the host machine that has virtual machines provisioned for IBM® Rational® Performance Tester might contain other virtual machines that use same set of resources. This scenario might also affect the measurement of response time.
To verify the response time accuracy of the schedule, run a very few number of virtual users on a control agent that is at another location. After the run, you can compare reports to verify the response time accuracy. If the results are comparable, perhaps within 10% of each other, you can be fairly confident that the response times reported by the load agents are accurate.
If you do not have a physical hardware to use the control agent approach, you can use the Run a Fixed Delay Measurement Audit on every location option. With this approach, you add one virtual tester on an agent to execute a delay action of perhaps 800 milliseconds. After the schedule run completes, you can measure the length of the schedule run with the applied delay to ensure response time accuracy. If the delay time is more than 1 second, you probably cannot rely on the response times being accurate from the virtual testers applying load on that agent.