Deep Kronecker ResNeXt forward fractional network-driven clustering for enhanced load balancing in cloud computing

Deep Kronecker ResNeXt forward fractional network-driven clustering for enhanced load balancing in cloud computing

연구 분야: Cryptography

학회: The Journal of Supercomputing

Load balancing is the crucial components in cloud computing, which make sure of the effective resource utilization, system reliability, and uninterrupted services. The effectual load balancing is the main challenge in cloud computing that requires advanced techniques to maintain Quality of Service (QoS). In this article, we develop the deep Kronecker ResNeXt forward fractional network (DK-ResNeXt FF-Net) model for this problem. This model simulates a cloud environment, which consist of multiple physical machines (PMs), virtual machines (VMs), a cloud manager, and a service provider. First, the round-robin strategy is employed for task distribution through VMs. After that, the classification of VMs is done by using the deep fuzzy clustering (DFC) technique, based on parameters, such as central processing unit (CPU) utilization, memory utilization, frequency scaling factor, million instructions per second (MIPS), and VM bandwidth. The gated recurrent unit (GRU) is employed for accurate load prediction and overload VMs relocate the tasks to underloaded VMs using the DK-ResNeXt FF-Net, which is formed by the integration of deep Kronecker networks (DKN), ResNeXt, and fractional calculus (FC). Moreover, the key parameters, such as scalability, response time, predicted load using GRU, fault tolerance, and QoS, are considered for task reallocation. When compared to other traditional models, the devised DK-ResNeXt FF-Net model achieved a minimum load of 0.072, resource utilization rate of 0.915, and response time of 8.087 s, for Setup-3 (18 PM and 26 VM). The proposed DK-ResNeXt FF-Net model provides significant improvements in load, resource utilization, and response time in cloud computing environments. By using advanced approaches such as DKN, ResNeXt, and FC, the designed model optimizes the redistribution of tasks among VMs, effectively addressing overload scenarios.