摘要
容器云作为一种高效、易维护和低成本的云环境解决方案,弥补了传统基于虚拟机的云环境构建方式的不足,受到国内外软件服务提供商的广泛关注。针对容器云基于阈值的响应式伸缩策略存在时间滞后性、难以及时响应服务资源请求的问题,提出一种基于二次移动平均法的预测式容器云伸缩方法。该方法在检测实时工作负载的同时通过二次移动平均法对未来工作负载进行预测,然后基于实时负载值和预测负载值利用响应式伸缩策略进行容器云伸缩决策。与原来基于阈值的响应式伸缩策略相比,在Docker swarm集群环境并施加周期性负载的实验条件下,提出的优化方法能够提前预测负载变化并调整集群规模,有效应对负载波动,服务响应时间波动幅度降低了约42.9%,保证了容器云中应用服务的质量和稳定性。
As an efficient,easy-to-maintain and low-cost cloud environment solution,container cloud makes up for the shortcomings of traditional virtual machine-based cloud environment construction,which has been widely concerned by software service providers at home and abroad.Aiming at the problem that the threshold-based responsive scaling strategy of container cloud has time lag and difficulty in responding to service resource request in time,we propose a predictive container cloud scaling method based on double moving average method.The method predicts the future workload by means of the double moving average method while detecting the real-time workload,and then uses the responsive scaling strategy to make the container cloud scaling decision based on the real-time workload value and the predicted workload value.Compared with the original threshold-based responsive scaling strategy,under the conditions of Docker swarm cluster environment and periodic workload,the proposed optimization method can predict the workload change in advance and adjust the cluster scale to deal with the workload fluctuation effectively.The fluctuation range of service response time is reduced by 42.9%,which ensures the quality and stability of service in container cloud.
作者
刘钱超
吴利
郑礼辉
LIU Qian-chao;WU Li;ZHENG Li-hui(Jiangnan Institute of Computing Technology,Wuxi 214083,China)
出处
《计算机技术与发展》
2019年第10期15-20,共6页
Computer Technology and Development
基金
国家重点研发计划(2016YFB1000505)
