EDLW-S: Efficient Deep Learning-based Workflow Scheduling for Rapid Cloud Computing

The rapid expansion of 5G mobile communications has led to an explosion of multimedia data traffic and reception, posing a significant challenge to cloud computing systems to process and manage data over limited bandwidth efficiently. Workflow scheduling includes efficient task scheduling and resource allocation, and is a key challenge in cloud computing for real-time multimedia applications. Traditional approaches often fail to address cost constraints, processing delays, and the priority of tasks. This work proposes a workflow scheduling framework based on deep learning for 5G cloud environments. It is a four-step process. Firstly, we improve the quality of multimedia data using adaptive histogram matching (AHE), which addresses the problems of spatial data formats. Then, we use the Gabor transformation to generate a representation of the time-frequency spread of the data. Feature extraction is then performed using statistical moments (SM), local binary patterns (LBP), and local optimal estimates (LOOP). Extractive functions are optimized using generalized normal distribution optimization (GNDO). Finally, FrameNet custom Convolutional Neural Network (CNN) classifies optimized functions to allow efficient prioritization and scheduling of tasks. The proposed solution, the LW-WSOA Algorithm, demonstrates significant improvements over baseline approaches in terms of makespan and execution cost, achieving an 87.28% reduction in makespan and an 88.46% reduction in execution cost. Experimental simulations confirm that LW-WSOA optimizes resource allocation, maintains high accuracy, and improves overall efficiency in 5 G-enabled cloud systems. In addition, ANOVA validates the statistical significance of said improvements, further emphasizing the algorithm’s efficacy in real-time multimedia applications requiring rapid processing and high throughput in next-generation 5G environments.