A high availability system based on the CNN-LSTM and volume replication algorithm
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.
万方数据
维普
