A Graph Neural Network-Based Approach with Dynamic Multi-Queue Optimization Scheduling (DMQOS) For Efficient Fault Tolerance and Load Balancing in Cloud Computing
Currently, cloud computing is expanding on a daily basis and has evolved into an efficient and adaptable paradigm for addressing large-scale issues. It is recognized as an internet-based computing model where various cloud users share computing and virtual resources like services, applications, storage, servers, and networks. Due to the rapid increase in the number of cloud users and their requests, cloud systems can be either under loaded or overloaded. These scenarios result in various issues, including prolonged response times and higher power consumption. In this study, we propose a novel approach that integrates Graph Neural Networks (GNNs) with Dynamic Multi-Queue Optimization Scheduling (DMQOS) to enhance the fault tolerance and load balancing capabilities of cloud computing environments. This paper presents a novel approach that leverages Graph Neural Networks (GNNs) in conjunction with DMQOS to efficiently address these challenges. Our approach, GNN-DMQS, focuses on adapting to the dynamic nature of cloud workloads by incorporating a DMQOS mechanism. By analysing this graph-based representation, the system can dynamically optimize the allocation of resources, facilitating effective load balancing while proactively mitigating potential faults. This dynamic approach optimizes resource utilization and response times, contributing to improve load balancing and system efficiency. Simultaneously, a GNN-based framework is employed to model and analyse the intricate dependencies among cloud resources and their performance. This GNN-driven approach enhances fault tolerance by predicting and mitigating potential failures in advance, thereby ensuring uninterrupted service availability. We evaluate the proposed approach, GNN-DMQOS, using extensive experiments on real-world cloud computing datasets. The results demonstrate significant improvements in both fault tolerance and load balancing compared to traditional methods. Moreover, our approach, GNN-DMQOS, exhibits adaptability to varying workloads, making it suitable for dynamic cloud environments.