Bandwidth abstraction and service rate instantiation for latency-bounded reliability provisioning in 5th generation wireless networks

Abstract

The provisioning of the statistical reliability regard to latency is crux for the bursty traffic in the 5th generation (5G) wireless network slicing. Based on martingale theory, a novel analysis framework of the latency-bounded reliability is proposed. For the queuing system with bursty traffic, a Queuing Length Martingale is constructed to reveal the impact of the entanglement between the arrival process and the service process on the latency performance, where these two stochastic processes are sharply heterogeneous. The complex martingale parameters are determined. Relying on the time shift feature of the Queuing Length Martingale, a Latency Martingale is defined, which enables to analyze the latency-bounded unreliability in the martingale domain. Leveraging the stopping time theory, a tight upper bound of the unreliability regard to latency is obtained. A bandwidth abstraction and service rate instantiation scheme is designed for wireless network slicing. Bandwidth abstraction facilitates to decouple the statistical reliability requirements as the desired bandwidth of the wireless node. In the 5G access network, a transmission power allocation problem is formulated with the constraint of abstracted bandwidth. According to the access states and channel characteristics, the demanded transmission power is captured dynamically to instantiate the service rates. Simulation results highlight the effectiveness of the bandwidth abstraction and service rate instantiation scheme.