门控时空卷积网络中的微服务时延预测模型  

Microservice latency prediction model based on gated spatial-temporal convolutional networks

作　　者：蒋溢[1] 冯啸林 杨川 熊安萍[1] JIANG Yi;FENG Xiaolin;YANG Chuan;XIONG Anping(School of Computer Science and Technology,Chongqing University of Posts and Telecommunications,Chongqing 400065,P.R.China;China Telecom Corporation Limited Luzhou Branch,Luzhou 646000,P.R.China)

机构地区：[1]重庆邮电大学计算机科学与技术学院,重庆400065 [2]中国电信股份有限公司泸州分公司,四川泸州646000

出　　处：《重庆邮电大学学报（自然科学版）》2024年第6期1175-1182,共8页Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)

基　　金：重庆市重点基金项目(2022NSCQ-LZX0128)。

摘　　要：微服务的多实例化部署和微服务间存在的依赖关系,使得准确捕捉微服务调用拓扑与微服务时延之间的关联性十分困难。针对该问题,提出了一种基于门控时空卷积网络(gated attention spatio-temporal graph convolutional network,GaSTGCN)的时延预测模型。通过构建微服务拓扑调用实例的图数据序列,采用多层时空卷积图神经网络模型,在考虑微服务调用拓扑的同时,捕捉微服务节点时间和空间上特征的关联性;考虑微服务应用规模日渐庞大、时空关联性更难捕捉的现状,结合门控卷积模型,采用膨胀卷积技术与自适应门控机制,更精确地获取局部与全局微服务依赖特征。实验表明,所提模型具有较好的收敛性能,并且预测精度优于传统的预测算法。The multi-instance deployment of microservices and their interdependencies make it challenging to accurately capture the relationship between microservice invocation topology and service delay.To address this issue,a delay prediction model based on the gated attention spatial-temporal graph convolutional network(GaSTGCN)is proposed.By constructing graph data sequences of microservice invocation instances,a multi-layer spatial-temporal graph convolutional network is employed to capture temporal and spatial feature correlations of microservice nodes while considering the service invocation topology.Recognizing the increasing complexity of microservice application scales and the difficulty in capturing spatial-temporal correlations,the model integrates gated convolution mechanisms with dilated convolution techniques and adaptive gating mechanisms to more accurately extract local and global microservice dependency features.Experimental results demonstrate that the proposed model achieves superior convergence performance and prediction accuracy compared to traditional prediction algorithms.

关 键 词：微服务 弹性伸缩 时延预测 门控时空卷积 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]