Recent Advances in Joint Resource Allocation, Security, and Efficient Task Scheduling in Cloud Computing Using Hybrid Pyramidal Convolution Split-Attention Networks: A Systematic Review

Abstract

Cloud computing has become a transformative technology that provides scalable resources, flexible storage, and on-demand services for modern applications. However, the rapid growth of cloud infrastructures has created major challenges in resource allocation, task scheduling, and data security. Large-scale cloud environments require dynamic management of heterogeneous resources such as CPU, memory, bandwidth, and storage while ensuring efficient task execution and protection against cyber threats. Traditional scheduling methods mainly focus on single objectives like execution time or cost and often fail to address security and complex resource interactions simultaneously. Recent advancements in deep learning have enabled intelligent cloud resource management through advanced optimization techniques. Attention-based convolutional neural networks and pyramidal convolution architectures can capture multi-scale workload features and improve scheduling adaptability. Hybrid models integrating pyramidal convolution and split-attention mechanisms help predict resource demands, detect anomalies, and optimize dynamic resource allocation while maintaining stability and security. Additionally, modern cloud systems must handle complex workloads generated by AI applications, IoT devices, big data analytics, and edge computing. Machine learning-driven scheduling frameworks improve load balancing, energy efficiency, latency reduction, SLA compliance, and operational cost management by optimizing workload distribution across virtual machines and data centres, thereby enhancing overall cloud computing performance and reliability.