Task Scheduling in Cloud Computing Using an Improved Genetic Algorithm

Task Scheduling in Cloud Computing Using an Improved Genetic Algorithm

reihaneh zia1

1) M.Sc. Student in Computer Engineering, Artificial Intelligence, Islamic Azad University, Borujen Branch, Iran Email:

IOI Code : ACCD-ACZKBA
Publication : 10th International Conference on Applied Researches in Science & Engineering (carseconf.com)
Abstract :
    Cloud computing has transformed task scheduling into a fundamental challenge in data center management by providing dynamic access to computational resources. The heterogeneity of virtual machines, variations in workload, and trade-offs among makespan, cost, and energy consumption limit the effectiveness of simple scheduling methods. This paper aims to propose a multi-objective task scheduling framework based on an improved genetic algorithm. The research methodology involves algorithm design and simulation-based empirical evaluation. In the proposed framework, each chromosome represents a mapping of tasks to virtual machines. The initial population is generated by combining heuristic-based solutions with random initialization. Parent selection is performed using tournament selection, while resource-aware crossover, adaptive mutation, and elitism are employed to enhance population diversity and preserve high-quality solutions. The fitness function combines makespan, cost, energy consumption, and load imbalance after normalization, using configurable weights. In addition, a repair mechanism is incorporated to correct invalid task assignments. For evaluation, the proposed algorithm is implemented in CloudSim and compared against the standard genetic algorithm, First-Come, First-Served (FCFS), Min-Min, and selected metaheuristic algorithms under consistent workload conditions. The evaluation metrics include makespan, deadline satisfaction, cost, energy consumption, resource utilization, and algorithm execution time. Analyzing multiple independent runs and reporting the mean, standard deviation, and statistical significance tests are intended to improve the reliability of the results. Furthermore, a sensitivity analysis of the algorithm s parameters is conducted. The adaptive and multi-objective mechanisms are expected to achieve a more favorable trade-off among execution speed, cost, and energy consumption. However, the extent of any improvement must be demonstrated through reproducible experiments. This framework can provide a foundation for developing scalable task schedulers for heterogeneous cloud computing environments.

Keywords : cloud computing task scheduling improved genetic algorithm multi objective optimization energy consumption