Fast Resource Allocation for Resilient Service Coordination in an NFV-Enabled Internet-of-Things System

Abstract

Network Functions Virtualization (NFV) is a new way of leveraging an Internet-of-Things (IoT) system to provide real-time and highly flexible service creation. In an NFV-enabled Internet-of-Things (NIoT) system, several IoT functions implemented as Virtual Network Functions can be linked as a service function chain to build a customized IoT service quickly. It is important for an IoT service to be able to recover from a failure. However, the supply of a resilient IoT service in an NIoT system is challenging due to the coordination of distributed VNF instances. In this paper, we formulate the problem of resilient service coordination in an NIoT system as a mixed-integer linear programming model, namely RSO\textsubscript{d}. The model offers the optimal resource allocation for minimizing service disruption when a failure happens at a node of an NIoT system. We also develop two modified versions of RSO\textsubscript{d} for different use cases required by an IoT provider. Further, two approximation algorithms are proposed to provide a resilient service for a large-scale NIoT system. The evaluation results show that RSO\textsubscript{d} and its modified versions produce the optimal resource allocation in significantly reduced time compared to previous work. The results suggest that an IoT provider should carefully select an appropriate resource allocation strategy as it has to pay a resource cost to minimize the service disruption. The results also show that our proposed priority-based heuristic algorithm outperforms an approximation algorithm based on Simulated Annealing in terms of the service disruption and computation time.