针对因果一致性模型约束过多且表达能力不足的问题,提出一种强的分布式上下文一致性模型.剔除同一用户的操作中非强制性的依赖关系,以及在不同客户端间定义所需的操作依赖;在此基础上,设计一种解决Paxos不能满足上层一致性需求问题的、实现分布式上下文一致性模型的共识算法CC-Paxos.利用时间戳给分布式上下文中的操作定序,采用细粒度的依赖检测,高效减少冲突操作的数目.实验结果表明,与在上层使用causal+一致性、下层用Egalitarian Paxos的方法相比,CC-Paxos显著降低了延迟,增加了吞吐量,且不需牺牲可扩展性.%Targeting at the inefficiency and insufficiency of the causal consistency model, a strong consistency model named distributed context consistency was proposed.Non-forced operational dependency was removed and the necessary dependencies among distributed clients were defined conveniently.On this basis, CC-Paxos was proposed to be the first consensus algorithm implementing the distributed context consistency model as well as keeping the reliability.Timestamps were exploited for operation sequencing in distributed contexts and fine-granularity dependency checking was adopted to effectively reduce the number of potential conflicts.Experimental results show that, compared with the method of using causal+ consistency model in the upper layer and Egalitarian Paxos in system layer, CC-Paxos can significantly decrease latency and increase throughput without sacrifice on scalability.
展开▼
