Scholarlink

<p>Multi-cloud execution of financial workflows requires the joint control of makespan, energy consumption, inter-cloud communication overhead, and privacy satisfaction. Existing schedulers usually optimize only part of those targets which always neglects that resources have heterogeneous DVFS states and tasks have various privacy levels. This paper presents DER, a differential evolution-based scheduler that combines (i) reference working-state initialization derived from DVFS-oriented energy minimization, (ii) privacy-feasible resource assignment, and (iii) a probabilistic hybrid mutation strategy integrating basic differential, target-to-best, multi-differential, and adaptive mutation operators. The scheduling problem is formulated as a four-objective optimization model that minimizes execution time, energy consumption, and cross-cloud transfer size while maximizing privacy benefit subject to deadline and privacy-feasibility constraints. Simulations on multi-cloud financial workloads compare DER with PSLS, SSPPH, and FFARS under low, medium, and high privacy loads. The results show that DER consistently yields fewer uncompleted tasks, lower execution time and cost, and higher privacy satisfaction, indicating that reference-state-guided initialization and adaptive mutation improve both convergence quality and feasibility preservation in privacy-aware financial scheduling.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

DER: energy-aware financial task scheduling in multi-cloud environments with privacy constraints

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月19日

查看全文
  • Xianmei Hua,
  • Liqun Yang,
  • Dai Jianhong,
  • Xinrong Zhan

摘要

Multi-cloud execution of financial workflows requires the joint control of makespan, energy consumption, inter-cloud communication overhead, and privacy satisfaction. Existing schedulers usually optimize only part of those targets which always neglects that resources have heterogeneous DVFS states and tasks have various privacy levels. This paper presents DER, a differential evolution-based scheduler that combines (i) reference working-state initialization derived from DVFS-oriented energy minimization, (ii) privacy-feasible resource assignment, and (iii) a probabilistic hybrid mutation strategy integrating basic differential, target-to-best, multi-differential, and adaptive mutation operators. The scheduling problem is formulated as a four-objective optimization model that minimizes execution time, energy consumption, and cross-cloud transfer size while maximizing privacy benefit subject to deadline and privacy-feasibility constraints. Simulations on multi-cloud financial workloads compare DER with PSLS, SSPPH, and FFARS under low, medium, and high privacy loads. The results show that DER consistently yields fewer uncompleted tasks, lower execution time and cost, and higher privacy satisfaction, indicating that reference-state-guided initialization and adaptive mutation improve both convergence quality and feasibility preservation in privacy-aware financial scheduling.