An experiment on lead-time impact in testing of distributed real-time systems

In the search for efficient test methods for distributed realtime systems, an experiment is conducted that investigates the impact on the lead-time re quired to identify a defect, depending on what information is available. On e group of experiment participants were given a system monitoring trace, wh ile a control group were given node monitoring traces. The information made available in the two sets of traces is identical except for which monitori ng approach was being used. The system under study is the control software for a private branch exchange (PBX) with three nodes. The experiment, condu cted with 23 participants, suggests with statistically significant confiden ce that it lakes shorter time to isolate a defect using system monitoring t races than using node monitoring traces. Further; the identification of the defect source is more correct with system monitoring. These findings help in motivating investments in technology that produces system monitoring tra ces.