Optimal execution tests and system reliability

Automatic routine execution (REx) testing of telecommunications switching equipment is often performed to improve system reliability, and to check the operational status of communications equipment. For duplex equipment-equipment composed of two identical units that needs only one unit working to function-REx testing, at a minimum, periodically checks the inactive unit for problems. If problems are found, then repairs can be made before service is impacted. In addition to this clear benefit, there are also costs associated with testing. Equipment that is being extensively tested may not be in-service, and the testing itself can lead to two types of incorrect conclusions, conclusions that lead to either unnecessary repairs, or to a dangerous misperception that the equipment is working properly. The first type of incorrect conclusion is a false positive-the REx test indicates there is a problem when there is none. The second is a false negative-the REx test indicates there is no problem when there is one. Since there are both costs and benefits associated with REx testing, it is natural to conclude that REx tests should be run neither too frequently nor too infrequently. For a variety of historical reasons, REx testing frequency is often performed on switch peripherals once a day. However, with changes in the failure rates of switching equipment, downtime requirements, mean time to restore equipment, and the variety of services offered by telephone companies, a daily REx testing frequency from a reliability point of view is not necessarily correct. In this paper we develop a detailed reliability model of a generic duplex peripheral module. The term generic is used, since the intent is to provide a useful methodology with some results that are independent of specific equipment, provided by various manufacturers. The model shows that picking the correct REx testing frequency results in a significant savings in downtime. The impact on system outages as a function of REx testing frequency is shown for a number of cases.