Dynamic Network Slice Assignment in Software-Defined IoT Networks

Network slicing using technologies such as software-defined networking is a promising approach to address the diverse quality-of-service (QoS) requirements of Internet of Things (IoT) applications. Network slicing allows a single physical network to be transparently divided into separate, logically independent networks, with distinct service levels per slice. This allows the network to support the diverse requirements of different IoT verticals ranging from connected vehicles to smart agriculture. In this paper, we propose a dynamic network slice assignment scheme for software-defined IoT networks, where each application type is assigned to a distinct slice controller depending on its QoS requirements. We formulate the optimal mapping between network slices and slice controllers as an integer programming problem, and show that the proposed scheme is capable of significantly reducing the number of QoS violations compared to existing controller mapping schemes. Further, we present a scheduling scheme to dynamically re-compute the optimal mapping, based on optimal stopping theory, while taking into consideration the temporal variations in traffic and link latencies. Simulation results show that the proposed scheduling strategy is able to reduce the number and expected cost of re-computations, while remaining within the QoS violation tolerance limit.