An Adaptive Observation Window for Verifying Configuration Changes in Self-Organizing Networks

The automatic verification of Configuration Management (CM) changes is an important step towards a highlyoptimized Self-Organizing Network (SON). A verification mechanism operates in three steps: based on the CM changes it divides the network into verification areas, assesses those by using an anomaly detection algorithm, and generates CM undo requests for the abnormally performing ones. To successfully fulfill those tasks, it has to sample the network for a certain time period, called an observation window. However, if the mechanism is timed improperly, it may generate too many false positive undo requests and may even prevent SON functions from reaching their set goals. To overcome this issue, we calculate for each cell a Cell Verification State Indicator (CVSI). It is based on the deviation from the expected performance and is updated using exponential smoothing. A verification area continuously reporting low CVSI values is considered as degraded and processed by the verification mechanism. The presented approach is evaluated in a simulated environment and compared it with other verification strategies. The results show that we are able to get a better result when we consider the CVSIs instead of the absolute performance values of the areas.