采用几何规划的多容器热迁移优化方法

Live Migration Optimization of Multiple Containers with Geometric Programming

容器热迁移是实现系统负载均衡,提高应用用户体验的一种有效方法。多容器热迁移需要解决总时间的优化问题,热迁移总时间包括停机时间和迁移时间。为实现多容器热迁移总时间的优化,本文提出了采用几何规划的多容器热迁移优化方法,采用预拷贝的迁移方案,提出迁移模型并结合几何规划方法进行问题求解。实验结果表明,内存拷贝轮数增加会减少停机时间和增加迁移时间;几何规划优化可以有效减少迁移总时间;增加停机时间的权重对迁移总时间的影响较大,仅需要少量内存拷贝轮数即可达到较为稳定的迁移总时间。

Container live migration is an effective way to achieve system load balance and improve user experience. The optimization of total migration time includes downtime and pre-copy time, and is the key problem needed to be solved in multi-container live migration. To optimize the total migration time of multi-container live migration, this paper proposes live migration optimization of multiple containers with geometric programming by adopting pre-copy algorithm, proposing migration model and optimizing with geometric programming. The simulation results show that increase of memory copy rounds can reduce downtime and increase migration time, and optimizing with geometric programming can effectively reduce total migration time, addition to the weight of down time has greater impact on total migration time, and only several memory copy rounds can reach a steady total migration time.