基于改进秃鹰搜索算法的Kubernetes资源调度应用

Improved Bald Eagle Search Algorithm for Kubernetes Resource Scheduling Application

Kubernetes是一个管理容器化应用的开源平台,其默认的调度算法在优选阶段仅把CPU和内存两种资源来作为计算节点的评分指标,同时还忽略了不同类型的Pod对节点资源的占用比例是不同的,容易导致某一资源达到性能瓶颈,从而造成节点对资源使用失衡.针对上述问题,本文在Kubernetes原有的资源指标基础上增加了带宽和磁盘容量,考虑到CPU、内存、带宽和磁盘容量这4类资源在节点上的占用比例对节点的性能的影响,可能造成Pod中应用的非正常运行,甚至杀死Pod,从而影响集群整体的高可靠性.本文将等待创建的Pod区分为可压缩消耗型、不可压缩消耗型以及均衡型,并为每种类型的Pod设置相应的权重,最后通过改进的秃鹰搜索算法(TBESK)来寻找出最优节点进行调度.实验结果表明,随着集群中Pod的数量在不断增加,在集群负载较大的情况下,TBESK算法的综合负载标准差和默认的调度算法相比提升了24%.

Kubernetes is an open-source platform for managing containerized applications.Nevertheless,its default scheduling algorithm only uses two resources,the central processing unit(CPU)and memory,as scoring metrics for computing nodes at the preference stage.Moreover,it also neglects the point that different types of Pods occupy different ratios of node resources.Consequently,a certain resource is highly likely to reach a performance bottleneck,ultimately causing an imbalance in the use of resources by nodes.To address the above problem,this study adds bandwidth and disk capacity to the original resource metrics of Kubernetes.The impact of the occupancy ratios of the four types of resources,i.e.,CPU,memory,bandwidth,and disk capacity,at the nodes on the performance of the nodes may cause abnormal operation of the applications in Pods and even kill the Pods,ultimately affecting the overall high reliability of the cluster.For this reason,the study classifies the Pods waiting to be created into the compressible consumption type,the incompressible consumption type,and the balanced type and sets corresponding weights for each type.Finally,the optimal nodes for scheduling are obtained by an improved bald eagle search algorithm(TBESK).The experimental results show that as the number of Pods in the cluster increases,the standard deviation of the synthetic load of the TBESK algorithm is 24%lower than that of the default scheduling algorithm in the case of a large cluster load.