Faulty-Trunk Detection Algorithms Using EADAS/ICUR Traffic Data

A message trunk, the basic connecting link in the switched telephone network, provides the communication path between switching machines as well as certain call setup capabilities, such as supervision, signaling, and ringing. For an important class of trunk faults that cause call failure, the trunk is released by the switching system upon customer abandonment and is again available to fail another call. As a result, a single undetected faulty trunk of this type can fail a significant fraction of the offered attempts to a group and will characteristically have an abnormally short holding time. Because of their potential service impact, significant efforts have been made to understand and quantify the impact that such abnormally 919 short-holding-time trunks have on central office and network service.1^1 It is now widely understood as a result of these studies that this generic trunk fault results in a fraction of service attempts "killed," which is out of all proportion to their number in the trunk population. Consequently, traffic data available from new and existing traffic data-acquisition systems has been viewed in the light of increasing interest in trunk-fault detection. In particular, with the advent of the Bell System EADAS/ICUR (Engineering and Administrative Data Acquistion System/Individual Circuit Usage Recorder) system,5 it was natural to ask whether the new individual trunk data available could be used to detect such "killer" trunks.* This paper discusses the theoretical aspects of a class of killer-trunk detection algorithms that utilize the individual trunk traffic data available in EADAS/ICUR.