容器云中基于改进遗传算法的资源分配策略

An Improved Genetic algorithm for resource allocation in container clouds

摘要:

容器很容易针对Web应用程序提供包装、迁移和配置等服务近年来已成为研究热点。提出了容器云中基于改进遗传算法的资源分配策略Double-GA。Double-GA是一种包括两个层次的资源分配策略：容器到虚拟机的资源分配和虚拟机到物理主机的资源分配。设计了容器云的两层资源分配的数学模型,以容器云中的整体物理主机能量消耗作为Double-GA策略的目标函数。Double-GA以遗传算法为基础,设计了双染色体的表达方式并处理好了遗传算法的初始化、进化、交叉、变异等操作。真实的实验实例数据结果表明：Double-GA双染色体算法明显优于普通遗传算法GA和递减最好适用算法Decreased Best Fit Algorithm。

Abstract:

Research of container is becoming a hot problem because it is easier for application providers to pack, migrate, and deploy web applications than using virtual machines. An improved Genetic algorithm for resource allocation in container clouds called Double-GA was proposed in this paper. Double-GA is a two-level resource allocation strategy, the containers are allocated to virtual machines and virtual machines are allocated to physical machines. The mathematics model of two-level resource allocation in container-based cloud was presented and the overall energy consumption of physical resource was designed as the objective function in GA. A new dual-chromosome representation was used for new genetic operators such as initialization, crossover, mutation and fitness function. The experimental results show that Double-GA gains much better than the single-GA and BFD algorithm in all test instances.