基于偏向稳定的分布式审计因果一致性模型

Distributed audit causal consistency model based on biased stability

在分布式存储中,因果一致性由于易编程与性能权衡最佳而备受青睐。为解决现有因果一致性成果中矢量依赖跟踪损失吞吐量的问题,提出了基于偏向稳定的分布式审计因果一致性模型。在查询操作中使用组合矢量时间戳替代全矢量时间戳,减少系统管理与通信开销。同时,将分布式关联数组引入因果审计,分区协同审计细化数据依赖性,以减少虚假依赖条目数量。理论分析与仿真结果表明,所提模型可提升吞吐量48.26%,降低更新响应时延16.25%。

In the distributed storage,causal consistency is favored due to the best trade-off between ease of programming and performance.To address the problem of vector-dependent tracking loss of throughput in existing causal consistency results,a distributed audit causal consistency model based on biased stability was proposed.Combined vector timestamps were used instead of full vector timestamps in query operations to reduce system management and communication overhead.Meanwhile,the causal auditing was introduced with the help of distributed associative arrays,and data dependency was refined by partitioned cooperative auditing to reduce the number of false dependency entries.Theoretical analysis and simulation results show that proposed model improves throughput by 48.26%and reduces update response latency by 16.25%.