集群环境中微服务容器资源特征分析及优化

在集群环境中部署微服务已经成为微服务部署的重要方式.由于不同种类服务对于CPU、内存、磁盘等资源的需求不同,导致集群中的节点产生资源碎片、出现资源消耗倾斜.如何提高集群资源利用率、降低集群能耗,成为继保障服务级别协议(service level agreement, SLA)之后的重大挑战.本文以阿里巴巴集团2021年发布的近两万个微服务的详细跟踪为数据样本,从容器资源使用情况、节点部署特征和资源消耗偏好等多个维度出发,分析其集群资源消耗特征,发现集群中出现了资源消耗倾斜的情况.通过进一步分析节点中容器部署情况发现容器资源分配不合理加剧了这一现象.基于此我们提出了一种使用深度双Q网络的模型,依据上游服务资源需求的实时变化,对容器资源分配进行优化.对比实验结果表明该方法可以在保证服务SLA的情况下有效提高容器资源利用率,改善节点资源消耗倾斜的情况.

Analysis and Optimization of Microservice Container Resources in Cluster Environment

Microservice deployment in a cluster environment has become an essential way. As different kinds of services have different demands on resources such as CPU, memory, and disk, it makes the nodes in the cluster produce resource fragments and become biased in resource consumption. How to enhance cluster resource utilization and reduce cluster energy consumption has become a major challenge after the service level agreement (SLA) is guaranteed. This study takes the detailed tracking of nearly 20000 microservices released by Alibaba Group in 2021 as data samples and analyzes their cluster resource consumption characteristics from multiple dimensions, such as container resource utilization, node deployment characteristics, and resource consumption preferences. As a result, the study finds that biased resource consumption occurs in the cluster. Further analysis of container deployment in the nodes reveals that this phenomenon is exacerbated by the inappropriate allocation of container resources. In view of this, this study proposes a model based on a deep dual-Q network to optimize container resource allocation according to the real-time changes of upstream service resource demands. The experimental results are compared, and it is shown that the method can effectively increase container resource utilization and ameliorate the bias of node resource consumption while serving SLA.