云计算环境下多租户软件优化调度仿真

Simulation on Multi-tenant Software Ooptimization Scheduling under Cloud Computing Environment

在传统的云计算环境中,多租户调度方式大都采用预分配机制下的静态资源调度方法,不能依据调度任务大小动态分配资源,容易出现"小任务大资源"问题,导致资源的严重浪费,调度效率大大降低。提出混合遗传算法与蚁群算法的多租户软件调度策略,将调度资源和调度任务离散成多个资源节点和任务节点,通过自然数对染色体进行编码,采用适应度函数评估染色体的质量,依据适应度值大小对遗传算法获取的子代个体进行排序,得到的最优秀个体当成初值,遗传算法获取的子代个体需要进行混乱整理,可以某个规则进行排序,之后,进行检索,获得的最优的个体供用户使用,满足要求时,停止,否则进行再次迭代。实验表明采用所提方法进行多租户软件调度,在运行时间、资源利用率以及用户数方面的性能都优于传统方法,具有较强的多租户软件调度能力,为多租户软件调度优化提供了科学依据。

In this paper, based on hybrid genetic algorithm and ant colony algorithm, a multi -tenant software scheduling strategy was proposed. Scheduling resources and scheduling tasks were dispersed into multiple resource nodes and task nodes respectively. The chromosome was coded by natural number, and then their qualities were evaluated by using the fitness function. According to the value of fitness, the offspring captured by genetic algorithm was ranked to get the best individuals which were used as initial values. On this process, the offspring can be sorted on one special rule, followed by the retrieval step. The optimal solution that meets the multi - tenant software scheduling accuracy was given for usage, meanwhile, the algorithm stopped when achieved the best software scheduling results. Otherwise, the local optimal solution was put into the parent individuals of genetic algorithm. Consequently, a new iteration was launched until obtaining the best multi - tenant software scheduling results. The results from experiments show that performances of this proposed method for multi - tenant scheduling are superior to traditional methods at run time, the efficiency of the resource utilization and the number of users. This method has strong capabilities for multi - tenant software scheduling and provides a scientific basis for the multi - tenant software scheduling optimization.