Optimizing Resource Allocation for Dynamic IoT Requests Using Network Function Virtualization

Abstract

Network Function Virtualization (NFV) is essential for ensuring efficient and scalable Internet-of-Things(IoT) networks. However, optimizing resource allocation in an NFV-enabled IoT (NIoT) system is challenging, particularly when IoT functions are distributed as Virtual Network Functions (VNFs). This paper presents an approach for optimizing function placement in a dynamic NIoT system deployed within a hierarchical edge cloud computing environment. We propose an integer linear programming model and approximation algorithms to maximize the number of satisfied requests while minimizing system costs for a given set of service requests. Additionally, we develop a deep reinforcement learning-based algorithm (RTL) to determine the optimal timing for relocating IoT functions as bandwidth requirements change. Our evaluation measures several key metrics, including deployment cost, end-to-end delay, and request acceptance ratio. The results demonstrate that the approximation algorithms achieve nearly optimal results in significantly less time. The RTL algorithm consistently improves operational costs across various traffic demand scenarios compared to a baseline algorithm. Furthermore, our findings suggest an investment strategy for NIoT service providers to enhance system performance and reduce costs.