基于键值存储的元数据集群副本一致性研究

设计一个高性能、高可靠性的分布式文件系统元数据集群,需要解决的一个很重要的问题便是副本一致性。传统的一致性维护方法大多用于保证副本间强一致性,但是强一致性容易对系统的读写性能造成较大影响。该文提出了一种基于键值存储引擎的元数据集群副本一致性管理策略,并从副本的放置、读写、恢复以及更新这4个方面展开详细阐述了具体的维护副本方法,最终实现灵活的、用户可配置的副本一致性约束。而且基于键值的存储方式能有效缓解负载均衡不足、数据可用性和查询性能欠缺的问题,满足元数据信息的高并发、高性能访问需求。

Replica Consistency Management Strategy of Metadata Cluster for Key-value Stores

To design a high performance and high reliability metadata cluster in distributed file system,needs to solve an important problem is replica consistency. Traditional consistency maintenance methods commonly used to ensure that strong consistency,but strong consistency is easy to impact system performance. This paper proposes a replica consistency management strategy of metadata cluster for key-value stores,and elaborate the maintain method in the aspects of replicas placement,read and write,restore and update, eventually achieve a flexible and user configurable replicas consistency constraint. Storage based on key-value can effectively relieve the load balance,data availability and query performance lack,satisfy the metadata information of high concurrency,high performance access requirements.