Bandwidth abstraction with the end-to-end latency-bounded reliability provisioning based on martingale theory

Abstract

The provisioning of reliability regard to end-to-end (E2E) latency is crux for Ultra-Reliable Low Latency Communications (URLLCs). Based on martingale theory and stochastic network calculus, an E2E latency-bounded reliability analysis framework is proposed for the multi-hop system, where the arrival traffic possesses burstiness and the services provided by nodes are heterogeneous. The Wald martingales of arrival and service processes are constructed respectively, which contribute to reveal the influence of multiple random processes entangled with each other on the E2E latency. Leveraging the Doob maximum inequality of martingales and the features of moment generation functions, a tight upper bound of the unreliability regard to E2E latency is captured. A martingale parameter, named as tandem service descriptor, is proposed to embody the features of the tandem service mode provided for the specifically bursty traffic. The bandwidth abstraction algorithm is designed, which decouples the E2E latency-bounded reliability requirement as the demanded bandwidth of each node. An instantiation scheme of service rates is proposed to conduct the dynamical transmission power allocation according to the access states and channel characteristics in the wireless access network.