Online SMFD training sets

Training sets for online SMFD. Accelerating networked data transfer is crucialfor enhancing the quality of cloud computing services. Thispaper establishes a new foundation for modern cloud cachingoptimization by integrating a predictive strategy, performancebounds, and analytical metrics into a comprehensive framework.Central to this work is the transformation of Shortest MaximumForward Distance (SMFD) offline approach, renowned for itstractable and near-optimal properties, into an online variantnamely ^SMFD. ^SMFD integrates a novelly resource-efficientmultilayer perceptron with dynamic input features to determinespatiotemporal characteristics of data usage, achieving superiorcost performances compared to both state-of-the-art and conventionalmethods, while maintaining computational simplicity.Additionally, we propose practical performance upper and lowerbounds that provide a rigorous basis for identifying performanceoptimization success and opportunities. These bounds also enablethe formulation of analytical metrics that support precise performanceevaluations, improve observability of algorithmic behavior,and reveal insights obscured by traditional metrics. Notably, thispaper represents the first endeavor to quantify performance gapsbetween these bounds and originally demonstrats that ^SMFDcan reduce these gaps by up to 15%.