摘要
Key-Value存储以其高扩展性和强一致性在如今热门的大数据时代中扮演着重要角色。而实现其扩展性重要的关键技术在于数据迁移机制的完善,数据迁移机制同时直接影响集群节点的负载均衡。Redis集群采用哈希算法将数据均匀分布在集群节点上,然而随着节点的不断扩展,Redis集群手动分配槽的迁移机制日益不足。针对上述问题,提出单位最大效益自适应迁移策略。一方面,负载自适应平衡策略通过实时判断集群负载情况解决Redis手动迁移的不足,有策略的数据迁移使其静态存储均衡化,从而实现系统之间的自适应调控;另一方面,单位最大效益迁移方案每次优先选择单位效益最大的节点进行迁移,直至迁移完成。和Redis性能相比,自适应迁移策略的迁移时间大概缩短了6.2%～14.1%,吞吐量约提高了4.9%～12.5%。
Nowadays,Key-Value storage has played an important role in the era of big data for its high extension and strong consistency.The key technology to achieve extensibility is to improve the data migration mechanism which also has affected load balancing of clusternode. Redis cluster utilizes hash algorithm to uniformly distribute data over cluster nodes. However,with the increasing of nodes,themechanism of manually allocating slots in Redis cluster becomes inadequate. In order to solve the above problems,we present an adaptivemigration strategy of unit maximum benefit. On the one hand,the strategy solves the shortage of manual migration of Redis by judgingload of cluster system real-time and achieves adaptive control between systems by strategic migration,which leads to static storage equilibrium. On the other hand,we propose a migration scheme of unit maximum benefit to optimize migration plan. The scheme will choosethe optimal node and find the best strategy until all migration are finished. Compared to Redis,the migration time of adaptive strategy hasshortened by 6. 2% - 14. 1% and the throughput has increased by 4. 9% - 12. 5%.
作者
韦立
陈珊珊
WEI Li;CHEN Shan-shan(School of Computer Science,Nanjing University of Posts and Telecommunications,Nanjing 210003,China;Jiangsu Key Laboratory of Big Data Security ＆Intelligent Processing,Nanjing 210003,China)
出处
《计算机技术与发展》
2018年第10期53-58,63,共7页
Computer Technology and Development
基金
国家自然科学基金(61572263)
