Praetorian: Probabilistic Planning for Radio Access Network Slice Assurance

Radio access network (RAN) slicing has been proposed to efficiently serve multiple services such as mobile broadband, augmented reality, gaming and low latency applications. The dynamic allocation of resources typically make use of physical resource blocks (PRBs) share configurations. However, there have been limited studies on the side-effects and uncertainty of dynamic changes in configurations. In this paper, we present Praetorian, a probabilistic planning approach to analyze RAN slice assurance. Detailed queuing network models are used to extract the probabilistic effects of domain actions. These are fed into probabilistic planners to select optimal actions such as priority changes, PRB share changes or dropping. Planning techniques are proposed that improve the resilience of slice assurance techniques under uncertainties. This approach is applied on a real use case from a mobile network operator requiring slice assurance.