摘要
由于微服务系统中调用关系的复杂性,目前广泛使用的基于门限值设定与监测的微服务自动伸缩方法由于负载瓶颈转移问题导致其所需时间较长。针对这一问题,提出基于负载关系图的微服务自动伸缩方法。该方法根据系统的历史调用链数据分析各个微服务之间的负载关系。在自动伸缩操作被触发后,通过请求频率的变化幅度以及微服务间负载关系计算各个微服务的伸缩需求,一并完成所有微服务的自动伸缩操作,从而减少自动伸缩所需时间。基于前述方法,设计并实现工具LRGAS并开展了自动伸缩方法对比实验。实验结果验证了基于负载关系图的微服务自动伸缩方法的高效性。
Due to the complexity of invocation relationship in microservice system,the currently widely used microservice auto-scaling method based on threshold setting and monitoring requires a longer time because of the problem of load bottleneck transfer.Aimed at this problem,a microservice auto-scaling method based on load relationship graph is proposed.The method analyzed the load relationship between each microservice according to the request trace data.After the auto-scaling operation was triggered,the auto-scaling requirements of each microservice were calculated according to the request frequency and load relationship between microservices,and then the auto-scaling operation of each microservices was completed in parallel,thereby reducing the time for auto-scaling.Based on this method,the tool LRGAS was designed and implemented while a comparative experiment of auto-scaling methods was carried out.The experiment result shows the efficiency of the microservice auto-scaling method based on the load relationship graph.
作者
冀超
彭鑫
赵文耘
Ji Chao;Peng Xin;Zhao Wenyun(School of Software,Fudan University,Shanghai 201203,China;Shanghai Key Laboratory of Data Science,Shanghai 201203,China)
出处
《计算机应用与软件》
北大核心
2022年第6期12-20,共9页
Computer Applications and Software
基金
国家重点研发计划项目(2018YFB1004803)。
关键词
微服务
调用链
自动伸缩
负载关系图
Microservice
Invocation chain
Auto-scaling
Load relationship graph
