分布式 NoSQL 系统写操作性能优化设计与实现

摘要

Distributed NoSQL systems have successfully demonstrated enormous potential.However, the write operations are not handled efficiently in mainstream distributed NoSQL systems, which account for a large proportion in applications.After analysing the existing problems first, we then present two optimised batching algorithms to handle the inefficiency problem of batch processing at service side.The prototype system is implemented based on a typical distributed NoSQL system, HBase.Compared with HBase officially released, this prototype can effectively raise batch processing efficiency of write operation and achieves up to 10x throughput, without sacrificing the consistency and durability of the previous system.%分布式NoSQL系统已经成功地展现出巨大的潜力。不过在很多应用场景中，主流分布式NoSQL系统对写操作的处理效率依然不高。首先对其中存在的问题进行分析，然后提出两个优化批处理算法解决服务端批量处理的低效问题。原型系统基于典型的分布式NoSQL系统HBase实现。与官方发布的HBase系统相比，在不影响原系统一致性和持久性设计的前提下，原型系统能有效提升写操作的批量处理效率，获得10倍的吞吐量。