Cloud Data Optimization: Lightweight Caching With Precise Bounds and Post-Cloud Metrics

Accelerating networked data transfer is crucial for enhancing the quality of cloud computing services. This paper establishes a new foundation for modern cloud caching optimization by integrating a predictive strategy, performance bounds, and analytical metrics into a comprehensive framework. Central to this work is the transformation of Shortest Maximum Forward Distance (SMFD) offline approach, renowned for its tractable and near-optimal properties, into an online variant namely <tex-math notation="LaTeX">SMFD</tex-math>. <tex-math notation="LaTeX">SMFD</tex-math> integrates a novelly resource-efficient multilayer perceptron with dynamic input features to determine spatiotemporal characteristics of data usage, achieving superior cost performances compared to both state-of-the-art and conventional methods, while maintaining computational simplicity. Additionally, we propose the determination of performance upper and lower bounds that provides a rigorous basis for identifying performance optimization success and opportunities. These bounds also enable the formulation of original metrics that support precise performance evaluations, improve observability of algorithmic behavior, and reveal insights obscured by traditional metrics. Notably, this paper represents the first endeavor to quantify performance gaps between these bounds and originally demonstrates that <tex-math notation="LaTeX">SMFD</tex-math> can reduce these gaps by up to 15%.