Temporal Graph based Overcommitted Routing for Deterministic Networking

Abstract

Deterministic networking technologies play significant roles in Internet of Things, Augmented Reality and so on. In fact, the resource reservation policy is employed to guarantee the time-deterministic transmission of flows in deterministic networks. However, for flows with dynamic bandwidth demands, this policy will allocate bandwidths according to the peak volume in rush hours, and the link bandwidths may be just authorized but not fully utilized, which results in bandwidth overprovision and low resource utilization. Besides, due to the time-varying characteristics of both the network topology and link bandwidths, traditional static graph based routing schemes are inefficient for time-varying networks. Therefore, it is important to find a feasible way to improve the bandwidth utilization while guaranteeing the deterministic performance of flows. In this paper, we explore the deterministic routing problem for flows with dynamic bandwidth demands in stochastic time-varying networks (STN). First, we adopt the stochastic time-varying graph to characterize the dynamic attributes of networks, including the link bandwidth and topology, where random variables are utilized to depict the dynamicness of both dynamic bandwidth demands of flows and time-varying bandwidth resources of links. Then, we design an overcommitment rule, which allows us to design a bandwidth resource overcommitment shortest path algorithm in STN (ROSP-STN) to guarantee the delay of flows. Finally, we run numerical experiments to evaluate performance of the ROSP schme. Simulation shows that the proposed ROSP scheme can increase the bandwidth utilization by 24% and the transmission delay of flows can also be guaranteed into a deterministic range.