摘要
云平台中的资源需求和负载情况会随着时间和用户需求发生变化,提高了算法平衡局部搜索和全局搜索的难度。因此,为提高云平台的吞吐量,提出基于改进模拟退火-禁忌搜索(Simulated Annealing-Tabu Search,SA-TS)算法的计算资源调度方法。首先,计算资源优先级,并将能耗最小、云服务成本最低和总时延最小作为目标,建立云平台计算资源调度目标函数;然后,采用模拟退火算法改进遗传算法的交叉机制,通过引入接受概率和逐渐降低温度的策略,使算法能够接受较差的解,从而跳出局部最优解,再遵循禁忌搜索思想优化遗传算法的变异机制,通过维护禁忌列表,避免重复搜索已知解,进一步增加跳出局部最优解的可能性。通过平衡局部搜索和全局搜索,使算法在解空间中更精细地搜索;最后,利用改进后的算法求解目标函数,完成对云平台计算资源的调度。仿真结果表明,SA-TS算法的收敛性好,可有效提高云平台的负载均衡性和平均吞吐量。
The resource requirements and load conditions in cloud platforms change with time and user demands,which increases the difficulty of algorithms in balancing local and global searches.Therefore,in order to improve the throughput of the cloud platform,a computational resource scheduling method based on the improved Simulated Annealing-Tabu Search(SA-TS)algorithm is proposed.First,the resource priority is calculated,and the objective function of computing resource scheduling for the cloud platform is established by taking the minimum energy consumption,the minimum cost of cloud service and the minimum total delay as the objectives.Then,the crossover mechanism of the genetic algorithm is improved by using the simulated annealing algorithm,and the algorithm is able to accept the worse solution by introducing the strategy of acceptance probability and gradually lowering the temperature,so that the algorithm can accept the poorer solution,thus jumping out of the local optimal solution,and then optimize the solution by following the idea of tabu search.The mutation mechanism of the genetic algorithm further increases the possibility of jumping out of the local optimal solution by maintaining the taboo list and avoiding repeated searches for known solutions.By balancing the local search and global search,the algorithm searches more finely in the solution space.Finally,the improved algorithm is utilized to solve the objective function and complete the scheduling of computing resources on the cloud plaform.The simulation results show that the SA-TS algorithm has good convergence and can effectively improve the load balance of the cloud platform and the average throughput of the cloud platform.
作者
刘超
梁雪青
袁兴佳
杜舒明
LIU Chao;LIANG Xue-qing;YUAN Xing-jia;DU Shu-ming(Guangzhou Power Supply Bureau of Guangdong Power Grid Co.,Ltd,Guangzhou Guangdong 510000,China)
出处
《计算机仿真》
2025年第4期435-438,450,共5页
Computer Simulation
基金
广州供电局项目(80010HK42210008)。
关键词
云平台
资源调度
模拟退火算法
遗传算法
禁忌搜索算法
Cloud platform
Resource scheduling
Simulated annealing algorithm
Genetic algorithm
Tabu search algorithm
