基于CNN-LSTM和卷复制方法的高可用系统设计方法  

A high availability system based on the CNN-LSTM and volume replication algorithm

作　　者：张焱[1] 李新建[2] 王畅 章建军[1] 陈小虎 邹鑫灏 严智 ZHANG Yan;LI Xinjian;WANG Chang;ZHANG Jianjun;CHEN Xiaohu;ZOU Xinhao;YAN Zhi(Informationization Technology Center,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China;China Tabacco Hubei Industrial LLC,Wuhan 430040,China;Wuhan Wenwen Information Technology Co.Ltd.,Wuhan 430040,China)

机构地区：[1]南京航空航天大学信息化技术中心,江苏南京210016 [2]湖北中烟工业有限责任公司,湖北武汉430040 [3]武汉问道信息技术有限公司,湖北武汉430040

出　　处：《南京邮电大学学报（自然科学版）》2024年第4期114-121,共8页Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition

基　　金：湖北中烟烟草关键应用科研委外项目(2022GLGL3XX2C077)资助项目。

摘　　要：针对单机服务器存在的单点故障问题,以及主备双机中存在的逻辑故障导致数据丢失的问题,设计了一种基于卷积和长短期记忆神经网络(CNN-LSTM)和卷复制方法的HA(High Availability)系统。系统至少包含两个节点,一个主节点以及一个或多个备用节点,主节点和备节点之间支持主备切换。每个服务器节点上包含4个模块,分别是负责接收配置信息与读写请求的代理模块;进行磁盘读写操作和重定向读写的磁盘I/O(输入输出)模块;负责主备节点间备份快照、映射表、数据块复制的卷复制模块以及基于CNN-LSTM进行状态检测的高可用模块。实验表明,该系统不仅可以解决单点故障问题,也可以解决主备双机集群中无法解决的逻辑错误问题;同时基于CNN-LSTM方法,自动针对服务器的运行健康状态进行分析和预测,可以根据预测结果自动通知管理员进行处理或自动进行主备切换。Since standalone servers usually face single-point failures and logical errors in master-slave dual machines can lead to data loss,this paper designs a high-availability(HA)system based on convolutional-neural-network-long-short-term-memory(CNN-LSTM)and volume replication technology.This system comprises at least two nodes:a primary node and one or more backup nodes.And it supports failover between the primary and the backup nodes.Each server node has four modules:a proxy module responsible for receiving configuration information and read/write requests;a disk IO module for disk read/write operations and redirecting reads/writes;a volume replication module for backup snapshots,mapping tables,and data block replication between the primary and the backup nodes;and a high availability module that utilizes the CNN-LSTM model to perform status detection.Experimental results demonstrate that this system can solve both the single point of failure and the logical error problems in primary-backup clusters,and can automatically analyze and predict server health status based on support vector machine methods.The prediction can be automatically send to administrators for manual handling,or the failover can be performed automatically.

关 键 词：卷复制 数据丢失 快照 CNN-LSTM 高可用系统 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]