Enhancing Cloud Resource Allocation with a Deep Learning-Based Framework

Abstract

Cloud computing has transformed the modern world of computing by enabling the provisioning of resources that are on demand and scalable. Nevertheless, the issue of efficient resource allocation persists due to unpredictable demand fluctuations and heavy duties. This paper investigates the allocation of cloud resources through the use of deep learning (DL) models, specifically Convolution Neural Networks (CNNs), Gated Recurrent Networks (GRNs), and Long Short Term Memory (LSTM) networks. CNNs acquire spatial patterns from cloud workload data, while GRNs acquire short-term resource usage dependency patterns. By acquiring long term patterns in work load variation, LSTMs further improve the accuracy of their predictions. Our proposed framework optimizes cloud resource allocation using these models, enhancing overall system performance, reducing energy consumption, and reducing latency. Experimental evidence indicates that our proposed deep learning framework is more precise and adaptable than conventional methodologies.