计算机研究与发展
計算機研究與髮展
계산궤연구여발전
JOURNAL OF COMPUTER RESEARCH AND DEVELOPMENT
2015年
7期
1531-1545
,共15页
周维%路劲%周可人%王世普%姚绍文
週維%路勁%週可人%王世普%姚紹文
주유%로경%주가인%왕세보%요소문
云计算%双层索引%并发跳表%范围查询%乐观并发控制
雲計算%雙層索引%併髮跳錶%範圍查詢%樂觀併髮控製
운계산%쌍층색인%병발도표%범위사순%악관병발공제
cloud computing%double-layer index%concurrent skiplist%range query%optimistic concurrency control
云数据处理在云计算基础设施中占有极其关键的地位。然而,当前的云存储系统绝大部分都采用基于分布式 Hash 的健‐值对模式来组织数据,在范围查询方面支持不理想、且动态实时性差,有必要构建云环境下辅助动态索引。通过总结、分析云环境中辅助双层索引机制,提出一种基于并发跳表的云数据处理双层索引架构。该架构采用两层体系结构,突破单台机器内存和硬盘的限制,从而扩展系统整体的索引范围。通过动态分裂算法解决局部服务器中的热点问题,保证索引结构整体的负载均衡。通过并发跳表来提高全局索引的承载性能,改善了全局索引的并发性,提高整体索引的吞吐率。实验结果表明,基于并发跳表的云数据处理双层索引架构能够有效支持单键查询和范围查询,具有较强的可扩展性和并发性,是一种高效的云存储辅助索引。
雲數據處理在雲計算基礎設施中佔有極其關鍵的地位。然而,噹前的雲存儲繫統絕大部分都採用基于分佈式 Hash 的健‐值對模式來組織數據,在範圍查詢方麵支持不理想、且動態實時性差,有必要構建雲環境下輔助動態索引。通過總結、分析雲環境中輔助雙層索引機製,提齣一種基于併髮跳錶的雲數據處理雙層索引架構。該架構採用兩層體繫結構,突破單檯機器內存和硬盤的限製,從而擴展繫統整體的索引範圍。通過動態分裂算法解決跼部服務器中的熱點問題,保證索引結構整體的負載均衡。通過併髮跳錶來提高全跼索引的承載性能,改善瞭全跼索引的併髮性,提高整體索引的吞吐率。實驗結果錶明,基于併髮跳錶的雲數據處理雙層索引架構能夠有效支持單鍵查詢和範圍查詢,具有較彊的可擴展性和併髮性,是一種高效的雲存儲輔助索引。
운수거처리재운계산기출설시중점유겁기관건적지위。연이,당전적운존저계통절대부분도채용기우분포식 Hash 적건‐치대모식래조직수거,재범위사순방면지지불이상、차동태실시성차,유필요구건운배경하보조동태색인。통과총결、분석운배경중보조쌍층색인궤제,제출일충기우병발도표적운수거처리쌍층색인가구。해가구채용량층체계결구,돌파단태궤기내존화경반적한제,종이확전계통정체적색인범위。통과동태분렬산법해결국부복무기중적열점문제,보증색인결구정체적부재균형。통과병발도표래제고전국색인적승재성능,개선료전국색인적병발성,제고정체색인적탄토솔。실험결과표명,기우병발도표적운수거처리쌍층색인가구능구유효지지단건사순화범위사순,구유교강적가확전성화병발성,시일충고효적운존저보조색인。
Cloud data processing plays an essential infrastructure in cloud systems .Without efficient structures ,cloud systems cannot support the necessary high throughput and provide services for millions of users .However ,most existing cloud storage systems generally adopt a distributed Hash table (DHT ) approach to index data , which lacks to support range‐query and dynamic real‐time character .It is necessary to generate a scalable ,dynamical and multi‐query functional index structure in cloud environment .Based on the summary and analysis of the double‐layer index systems for cloud storage ,this paper provides a novel concurrent skiplist based double‐layer index (referred as CSD‐index ) for cloud data processing . Two‐layer architecture ,which can breakthrough single machine memory and hard drive limitation ,is used to extend indexing scope .Online migration algorithm of skiplist’s nodes between local servers is used to make dynamic load‐balancing . The details of the design and the implement of the concurrent skiplist are discussed in this paper .Optimistic concurrency control (OCC) technique is introduced to enhance the concurrency .Through concurrent skiplist CSD‐index improves the load bearing capacity of the global index and enhances the overall throughput of the index .Experimental results show the efficiency of the concurrent skiplist based double‐layer index and it has viability as an alternative approach for cloud‐suitable data structures .