Energy- and Thermal-Aware Resource Management for Heterogeneous Edge–Cloud Systems with Workload-Dependent Capacity Degradation

This repository contains the fully specified reference configuration, retained numerical outputs, corrected statistical summaries, verification code, and figure-source files supporting the study “Energy- and Thermal-Aware Resource Management for Heterogeneous Edge–Cloud Systems with Workload-Dependent Capacity Degradation.” The study examines heterogeneous edge–cloud systems in which current workload affects subsequent service capability through thermal degradation and cooling. The RCV-1 reference configuration combines finite-batch BMAP arrivals, Erlang-2 phase-type service, heterogeneous high-performance and energy-efficient edge servers, first-order RC thermal dynamics, hysteretic cooling, thermal-aware admission, and elastic cloud fallback. The package includes: (1) dual-scale 1H+1E fine-grid CTMC versus continuous-RC Hybrid-DES validation; (2) the 2H+2E workload–thermal campaign; (3) M0/M1/M2 thermal-model simplification and policy screening; (4) equal-mean Poisson, Moderate, and Strong traffic families; (5) a complete 63-policy enumeration with marginal and policy-family screens; (6) heterogeneous-server composition comparisons and architecture-specific searches; (7) cloud-energy crossover bootstrap intervals; and (8) high-resolution publication figures and their source data. Version 3 is the controlled corrected release aligned with the audited manuscript. It replaces the previous archive with corrected asymptotic customer-count dispersion indices for the Moderate and Strong BMAPs, exact continuous-time integration of fractional nominal-capacity erosion and the 75°C risk-reference fraction, threshold-split thermal-regime accounting, a source-derived six-candidate M0 selection procedure, and a joint familywise screen covering all 315 architecture–policy combinations. Under the declared familywise rule, policies 53 and 54 form the empirical sample-mean Pareto set, and only the 2H2E architecture has policies passing the joint architecture–policy screen among the five tested compositions. Independent verification code reconstructs the reported summaries, validation comparisons, traffic descriptors, erosion identities, policy selections, feasibility decisions, cloud-energy crossover values, Pareto membership, architecture diagnostics, figure-source values, and SHA-256 manifest. All 750 corrected archived-result verification checks pass with zero failures. No external empirical dataset is used. The experiments are based on analytically specified stochastic models and synthetic, mechanism-oriented numerical configurations. The archive supports verification from retained outputs and provides executable code for the workload–thermal campaign. It does not claim platform-specific thermal calibration or provide a single end-to-end simulator for regenerating every experiment.本代码仓库包含支撑研究《面向存在负载依赖型容量退化的异构边缘-云（Edge–Cloud）系统的能效与热感知资源管理》的完整规范参考配置、留存数值输出结果、修正后的统计汇总、验证代码以及图像源文件。 本研究聚焦异构边缘-云系统，其中当前负载会通过热退化与冷却过程影响后续服务能力。RCV-1参考配置整合了有限批量马尔可夫到达过程（Batch Markovian Arrival Process，BMAP）到达、爱尔朗-2（Erlang-2）相位型服务、异构高性能与高能效边缘服务器、一阶RC热动力学、迟滞冷却、热感知准入控制以及弹性云回退机制。 本套件包含：(1) 双尺度1H+1E精细网格连续时间马尔可夫链（Continuous-Time Markov Chain，CTMC）与连续RC混合离散事件仿真（Hybrid Discrete Event Simulation，Hybrid-DES）的验证实验；(2) 2H+2E负载-热特性实验集；(3) M0/M1/M2热模型简化与策略筛选；(4) 等均值泊松、中等（Moderate）与强（Strong）流量族；(5) 完整的63种策略枚举，附带边际筛选与策略族筛选；(6) 异构服务器组合对比与架构专属搜索；(7) 云能交叉点自举置信区间；(8) 高分辨率出版级图像及其源数据。 版本3为与经审核手稿对齐的受控修正发布版本。相较于此前的存档，本版本修正了中等与强BMAP的渐进式客户数离散指数，实现了分数标称容量侵蚀与75℃风险参考占比的精确连续时间积分，新增了阈值分割热状态统计、源自源码的6候选M0选择流程，以及覆盖全部315种架构-策略组合的联合族级筛选机制。依据声明的族级规则，策略53与54构成了经验样本均值帕累托（Pareto）集；在5种测试的服务器组合中，仅2H2E架构的策略通过了联合架构-策略筛选。 独立验证代码可复现报告中的汇总结果、验证对比、流量描述符、容量侵蚀标识、策略选择、可行性判定、云能交叉点数值、帕累托集成员资格、架构诊断、图像源数值以及SHA-256清单。全部750项修正后的存档结果验证检查均通过，无任何失败案例。 本数据集未使用外部经验数据集。所有实验均基于解析定义的随机模型与面向机制的合成数值配置。本存档可通过留存输出结果支持验证，并为负载-热特性实验集提供可执行代码。本存档不提供专属平台的热校准方案，也未提供可复现所有实验的端到端单一模拟器。