UE Blocking Probability Model for Planning 5G Guaranteed Bit Rate (GBR) RAN Slices

Abstract

Modeling the probability of blocking User Equipment (UE) sessions is key for planning in advance the amount of radio resources required by a Guaranteed Bit Rate (GBR) slice. This task is challenging since the amount of these resources depends on factors such as the channel quality of each UE, the packet scheduler discipline, and the GBR requirements. Under this context, we propose an analytical model to evaluate the UE blocking probability for a GBR slice. A key aspect of the proposed model is considering as input the distribution of the average Signal-to-Interference-plus-Noise Ratio (SINR) experienced by the UE, therefore allowing to capture the radio conditions in the cell. Our model builds a multi-dimensional Erlang-B system which meets the reversibility property. This involves our model is insensitive to the holding time distribution for the UE sessions. The reversibility property also involves the state transition probabilities have product form, so that the computational complexity of our model is low. Furthermore, the proposed model captures the SINR gain provided when the base station implements a channel-aware packet scheduler to allocate radio resources to the UE sessions. We validate the proposed model, demonstrating an estimation error for the UE blocking probability below 1.5%.