一种基于数据状态同步的一致性策略研究

Research on Consistency Strategy Based on Data State Synchronization

在传统的分布式系统中,修改数据后,通过数据同步复制的方式可以使多个节点保持数据的访问一致性,同时对主节点加锁,可以保证数据的写入一致性。但数据同步的方式会造成达到访问一致性的时间长、数据写入性能低。本文提出了一种基于数据状态同步的一致性策略,能够加快同步时间,提升写入性能。通过结合Zookeeper分布式协调服务设计了一种仅同步数据信息,完成对过期数据块的注销,从而保证数据的访问一致性,同时设计了分布式锁来确保数据写入的一致性。实验表明本文方法具有较低的同步时长,且并发写入性能更强。

In a traditional distributed system,after data is modified,multiple nodes can maintain data access consistency through data synchronization replication,and at the same time lock the master node to ensure data write consistency.However,the method of data synchronization will cause a long time to reach access consistency and low data writing performance.This paper proposes a consistency strategy based on data state synchronization,which can speed up synchronization time and improve write performance.By combining with the Zookeeper distributed coordination service,a system that only synchronizes data information is designed to complete the cancellation of expired data blocks,thereby ensuring the consistency of data access,and at the same time designing a distributed lock to ensure the consistency of data writing.Experiments show that the method in this paper has a lower synchronization duration and a stronger concurrent write performance.