Survey of Graph Neural Network for Dynamic Resource Allocation in SDN-Enabled Edge Computing and IoT Networks: A Review

Abstract

The present review article discusses the most recent applications of the Graph Neural Networks (GNNs) in addressing the problem of resource distribution in Software-Defined Networking systems (SDNs) combining edge computing and IoT. The analyzed studies depict how GNNs are intelligent at capturing complex interactions within networks based on the changes in data flow and device demands. Unlike traditional methods of distributing resources, GNN-based methods have a better level of performance due to the ability to perceive spatial associations between network components and temporal consumption patterns. They demonstrate promising results of reduction in latency, refined bandwidth allocation and enhancement in energy consumption with continued focus on existing problems involving scalability, real time processing that require further input. The review highlights the importance of some key insights that demonstrate the ability of GNNs to assist in solving major issues in resource allocation. An example illustrating this is the use of GNN-enabled methods to minimize the latency due to the intelligent way traffic is redirected through the most efficient routes hence improving the overall agility of the network. Additionally, these tools have worked well to optimize daily bandwidth distribution such that the resources are served to boost throughput and reduce congestion. Also, GNNs have become critical in enhancing the energy efficiency which is a critical consideration to the IoT and edge computing environments where devices are limited by resources; the ion of energy is very stringent. GNNs can help in extending the working life of equipment that is sensitive to energy by adjusting the resource allocation based on real-time information.