SPARF research package: code, locked evaluation protocol, predictions, and results
Abstract
Research package supporting the article "SPARF: Seasonal Profile-Assisted Ridge Forecasting for Proactive Resource Hotspot Prediction in Edge Computing" (J.-P. Yang), prepared for submission to Computer Networks. It contains the experiment programs, the locked evaluation protocol, VM roles and data hashes, saved baseline models, per-origin predictions, coefficients, formal SPARF evaluation results, same-run latency measurements, the supplementary test comparison, audit records, and figures. recompute_metrics.py reproduces the reported metrics, bootstrap intervals, and warning coverage from the saved predictions without the raw traces or retraining; verify_package.py checks file integrity against SHA256SUMS. The raw GWA-T-13 Materna traces are not redistributed; they are available from the Grid Workloads Archive and were provided by Materna GmbH Information & Communications, Dortmund, Germany.