Abstract:

The traditional Ant Colony Optimization （ACO） algorithm in the process of cloud computing resource scheduling has the defects of unreasonable task and resource node matching， task execution with long time and high cost， unbalanced virtual machine load， and low execution efficiency of the cloud computing system. To address these problems， Ant Colony-Simulated Annealing Algorithm （AC-SAA） was proposed， which aimed to reduce the task execution cost， shorten the task execution time， and keep the system load balanced， and establish the fitness functions of task execution time， cost， and load balancing rate to improve the heuristic factor of the traditional ACO algorithm. The locally optimal solution was solved by the ACO algorithm， and then the solution was further optimized and the pheromone was updated using the simulated annealing algorithm to obtain the globally optimal solution. The proposed algorithm achieved a reasonable allocation of cloud resource nodes and tasks and accelerated the convergence of the algorithm. Experimental results show that compared with the traditional ACO algorithm， AC-SAA shortens the iteration times by at least 52.2%.

Key words: cloud computing, Ant Colony Optimization (ACO) algorithm, simulated annealing algorithm, pheromone, task scheduling

摘要：

传统蚁群优化（ACO）算法在云计算资源调度过程中存在任务与资源节点匹配不合理、执行任务时间长且成本高、虚拟机负载不均衡和云计算系统执行效率低等问题。针对这些问题，提出一种蚁群模拟退火算法（AC-SAA），该算法以降低任务执行成本、缩短任务执行时间、保持系统负载均衡为目标，建立任务执行时间、成本、负载均衡率的适应度函数改进ACO算法的启发因子。通过ACO算法求解出局部最优解，然后利用模拟退火算法对该解进一步优化和信息素更新，从而获取全局最优解。所提算法实现了云资源节点与任务的合理分配，加速了算法收敛。实验结果表明，AC-SAA比传统ACO算法在迭代次数方面减少了52.2%以上。

关键词: 云计算, 蚁群优化算法, 模拟退火算法, 信息素, 任务调度