Stochastic Modeling of SLA-Aware Edge–Cloud Orchestration under Cold Starts
This repository contains the reproducibility data and verification code supporting the study “Stochastic Modeling of SLA-Aware Edge–Cloud Orchestration under Cold Starts: Deadline-Phase Scheduling and Elastic Provisioning.” The study develops a stochastic edge–cloud orchestration model that jointly represents observable deadline phases, marked-batch arrivals, SLA-aware service scheduling, non-cancellable stochastic cold starts, elastic-capacity activation, retention, release, and draining. Controlled synthetic experiments are used to isolate the effects of arrival correlation, deadline evolution, service speed, activation and release thresholds, and setup-time distributions. The package contains an independently executable explicit-customer discrete-event simulation, selected replication-level numerical outputs underlying the confirmatory sensitivity analyses, summary statistics and contrasts, exact CTMC–DES certification records for two tractable anchors, software requirements, execution instructions, automated verification code, and SHA-256 integrity checks. The deposited evidence comprises 1,560 replication-level observations and two exact CTMC–DES anchors covering 18 simultaneous metric comparisons. The synthetic experimental design is intentional. Production traces generally confound or omit the individual mechanisms required for controlled identification, including burst correlation, deadline-phase evolution, service-speed variation, cold-start distributions, and capacity-retention decisions. No external, proprietary, or confidential dataset is used. The automated verification script checks file integrity, metadata consistency, customer-flow conservation, activation balance, replication records, CTMC boundary and overflow gates, and simultaneous CTMC–DES validation results. The standalone DES can also be executed as an independent smoke test. The deposited CSV and JSON files constitute the authoritative numerical record for the reported confirmatory and certification results.