Threshold Pair Selection for the Reliable Condition Monitoring of Telecommunication Services

A service provider usually measures the conditions of its service(s) via Key Performance Indicators (KPIs) and lets a Condition Monitoring System (CMS) continuously oversee them. A CMS additionally notifies the responsible workforce of abnormal KPI values. Telecommunication services often show time-varying load characteristic, i.e. their amplitudes of use fluctuate time-dependently. On account of that, a CMS has to cope the Condition Monitoring (CM) of such telecommunication services by dint of fluctuating thresholds. A couple of existing disquisitions have pioneered the calculation of dynamical thresholds to this end. Aside from the normal condition, contemporary CMSs can typically classify into the two measurable abnormal states warning and critical by way of two collaborating thresholds. Up to now, nobody has publicized a mechanism to extract the two ideal thresholds from an array that comprises at least three available ones. This contribution closes this gap by introducing three threshold extraction strategies, scrutinizing their reliability, and recommending ideas for further ones.