Scholarlink

<p>Task scheduling in cloud computing environments is a complex, multi-objective optimization problem that requires balancing conflicting goals such as minimizing execution time and cost while maximizing resource utilization and throughput. Metaheuristic algorithms are effective for solving such problems because of their adaptive, stochastic, and balanced search behaviors. In this work, we propose a family of hybrid scheduling algorithms that combine the strengths of two well-established metaheuristics for cloud task scheduling: Fireworks Algorithm (FWA)—known for its diverse exploration of the search space, and Whale Optimization Algorithm (WOA)—recognized for its effectiveness in local exploitation. This study presents a novel framework that strategically combines two scheduling algorithms exhibiting complementary search behaviors. To achieve this, four hybridization strategies are developed, namely Sequential Hybrid, Parallel Hybrid, FWA-Encircling Hybrid, and WOA-Spark Hybrid, each formulated based on a specific design rationale. To the best of our knowledge, this is the first systematic study to explore multiple hybridization pathways between two metaheuristic algorithms for cloud task scheduling. A comprehensive experimental analysis executed on heterogeneous CloudSim Plus environments using both synthetic workloads and the real-world GoCJ trace reveals that the proposed hybrids consistently outperform standalone FWA and WOA, and recent advanced metaheuristic scheduling algorithms&#xa0;(including OBDFWA, GA–GWO, MWOA, and FireBat) across makespan, composite fitness (by 2.5%–7.5%), and CPU/RAM/bandwidth utilization, while remaining competitive in power consumption under varying task loads. We further evaluate the deployment feasibility of the proposed hybrids via scheduling overhead and algorithmic convergence analysis under increased workload sizes (up to 1000 cloudlets). Statistical validation using Friedman’s ranking test with Nemenyi post-hoc comparisons further confirms the statistical significance and reliability of the observed improvements.</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

Fusing exploration and exploitation: hybrid fireworks–whale optimization for multi-objective independent task scheduling in cloud environments

  • 期刊论文

  • 发表日期：

    2026年7月23日

查看全文
  • Abdullah Nayem Wasi Emran,
  • Majisha Jahan Disha,
  • Rezwana Reaz

摘要

Task scheduling in cloud computing environments is a complex, multi-objective optimization problem that requires balancing conflicting goals such as minimizing execution time and cost while maximizing resource utilization and throughput. Metaheuristic algorithms are effective for solving such problems because of their adaptive, stochastic, and balanced search behaviors. In this work, we propose a family of hybrid scheduling algorithms that combine the strengths of two well-established metaheuristics for cloud task scheduling: Fireworks Algorithm (FWA)—known for its diverse exploration of the search space, and Whale Optimization Algorithm (WOA)—recognized for its effectiveness in local exploitation. This study presents a novel framework that strategically combines two scheduling algorithms exhibiting complementary search behaviors. To achieve this, four hybridization strategies are developed, namely Sequential Hybrid, Parallel Hybrid, FWA-Encircling Hybrid, and WOA-Spark Hybrid, each formulated based on a specific design rationale. To the best of our knowledge, this is the first systematic study to explore multiple hybridization pathways between two metaheuristic algorithms for cloud task scheduling. A comprehensive experimental analysis executed on heterogeneous CloudSim Plus environments using both synthetic workloads and the real-world GoCJ trace reveals that the proposed hybrids consistently outperform standalone FWA and WOA, and recent advanced metaheuristic scheduling algorithms (including OBDFWA, GA–GWO, MWOA, and FireBat) across makespan, composite fitness (by 2.5%–7.5%), and CPU/RAM/bandwidth utilization, while remaining competitive in power consumption under varying task loads. We further evaluate the deployment feasibility of the proposed hybrids via scheduling overhead and algorithmic convergence analysis under increased workload sizes (up to 1000 cloudlets). Statistical validation using Friedman’s ranking test with Nemenyi post-hoc comparisons further confirms the statistical significance and reliability of the observed improvements.