Network testing [Series Editorial]

The geographical scale of network testing depends on the size of the testbed. It could be as small as the desk area in a laboratory, medium on a campus or within a building, or as large as the wide area in the Internet. Traditionally, one would construct the testbed in a laboratory so that all the environmental parameters could be programmed and controlled, and the results could be reproduced. However, since some of the environmental parameters are too complicated to program and control, often we need to use the real environment as the testbed. This extends the testbed from a laboratory to a building, a campus, or even the Internet. As the parameters in the real environment are often not programmable and controllable, we thus lose the exact reproducibility of results. One common way to solve this problem is to repeat the experiments for a prolonged period and average over the huge number of repetitive runs. Then the testing period is usually weeks to months instead of hours. Another less common way is to "capture" the environmental parameters as traces and replay them onto a laboratory testbed. Precise reproducibility can be achieved, but the environment parameters in the captured traces can only be ???replayed??? instead of programmed and controlled. The results from such tracedriven testing are useful if the environments and captured traces are representative enough.