38 11 2013 11 GeomaticsandInformationScienceofWuhanUniversity Vol.38No.11 Nov.2013 :1671-8860(2013)11-1339-05 :A GIS Key-value 1 1 1 1 (1 129 430079) : 设计了一种基于 Key-value 结构的缓存 KV-cache 旨在简化数据结构 高效管理缓存数据 KV-cache 由 Mem-cache 和 F-cache 构成其中 Mem-cache 为第一级缓存 并采用缓存回写机制以异步方式将数据写 入 F-cache 进行持久化存储 F-cache 使用大文件存储数据 并构建了基于 B 树的缓存索引以加快检索速度 数据写入 F-cache 的位置由基于 B 树的空闲空间管理器进行分配在 100 Mbps 局域网内测试 KV-cache 读写性能实验表明 KV-cache 读写性能优于传统文件目录方式的缓存 能够安全 高效地管理所缓存的各种 空间数据 : 分布式 GIS; 缓存 ;KV-cache;Key-value 存储 ; 空闲空间管理 :P208;TP311.13 GIS (Key) (Value) Value GIS ;Key [1-2] Key-value [3-5] GIS [11-12] Key-value Key-value GIS Bigtable [12] Dynamo [11] Cas- sandra [13] Key-value GIS [3-9] NASA World Wind [410] 2.1 I/O (KV-cache) Key- [4] value Key-value GIS DEM KV-cache 1 Key-value 2 (Mem-cache) (F-cache) Mem- cache F-cache F-cache ;F-cache Key-value :2013-07-08 : (201001046); (2011BAH16B08)
1340 2013 11 1 KV-cache 1 KV-cache Fig.1 WorkflowofKV-cache 2.2 Key-value Key ; Key Mem-cache (FIFO) Value Value Key Mem-cache ( Value 256B 100 MB 20%) F- cache Key-value 2.4 (F-cache) Key Value F-cache 3 2 2 64 F-cache ; 8 ; 64 ; Value 3 F-cache Fig.3 SchematicDiagramofF-cacheStructure B Oracle MySQL B GB 2.3 (Mem-cache) Mem-cache F-cache F- cache I/O Mem-cache B : Key Mem-cache Key Mem-cache F-cache 2 Fig.2 ASampleofCachingImagePyramid
38 11 : GIS Key-value 1341 WindowsXP 90 m F- SRTM 10 cache 512 512 jpg 700975 8.57GB F-cache 5~154 KB 5 3 64 B 4 F-cache ; 5 SRTM ; Fig.5 RenderingSRTMImagePyramidintheClient C# ( ) KV- cache F-cache 1GB 5minKV-cache Mem-cache KV-cache 4 F-cache ; Mem-cache Fig.4 SchematicDiagramofWritingCache KV-cache BlocktoF-cacheUnderSeveralConditions F-cache F-cache 2 3 SRTM 50 MB 6 KV-cache F- cache 6(a) KV-cache F-cache 5min 10min 15min 20min 25min 30min Mem-cacheKV-cache F-cache 6(b) C# 100Mbps 6 C/S Fig.6 AverageRead/WriteTimeofDiferentCaches :CPU AMDPhenom IX49554 GB 1TB WindowsServ- er2003; :CPU Intel Dual E22002 GB 500 GB 7 KV- cache 7(a) Mem- cache
1342 2013 11 KV-cache 15min KV-cache 5min KV-cache 2min ; 15 min GIS KV-cache 7(b) Mem-cache [1]. [M].2. : 2010 7 KV-cache 20094:55-59 Fig.7 AverageRead/WriteElapsedTimeofKV-cache withdiferenttimingintervals 200530(12):1092-1095 8 KV-cache Mem-cache [6]. 8(a) 5 [J]. ( min Mem-cache )20099:37-39 [7] YangC WWong W DYangR Xetal.Perform- Mem-cache ance-improvingtechniquesin Web-based GIS[J]. InternationalJournalof GeographicalInformation F-cache Science200519(3):319-342 [8]. P2P 8(b) [J]. 20093:236- Mem- 241 cache KV-cache [9]. WebGIS [J]. 20042:177-185 Fig.8 8 4 Mem-cache KV-cache AverageRead/WriteTimeofKV-cachein DiferentMem-cacheCapacities StructuredStorageSystem[J].ACM SIGOPS Op- DPGrid C/S eratingsystemsreview201044(2):35-40 DPGrid : KV-cache E-mail:tuzhenfa@126.com KV-cache [2]. (QoGIS): [J]. 200732(5):385-388 [3]. WebGIS [J]. 20075:41-44 [4]. 3 [J]. [5].WebGIS [J]. wind.arc.nasa.gov/index.html2011 [10]NASA.NASA World Wind[OL].htp://world- [11]DecandiaGHastorunDJampaniMetal.Dyna- mo:amazon s Highly Available Key-valueStore [J]. ACM SIGOPS Operating Systems Review 200741(6):205-220 [12]ChangFDeanJGhemawatSetal.Bigtable:A DistributedStorageSystemforStructuredData[J]. ACM TransactionsonComputerSystems(TOCS) 200826(2):1-26 [13]LakshmanAMalikP.Cassandra -A Decentralized GIS
38 11 : GIS Key-value 1343 Key-valueCachingofGeospatialDataforDistributedGIS TU Zhenfa 1 MENG Lingkui 1 ZHANG Wen 1 HUANG Changqing 1 (1 SchoolofRemoteSensingandInformationEngineeringWuhanUniversity129LuoyuRoadWuhan430079China) Aboutthefirstauthor:TUZhenfaPhDcandidatespecializedinspatialdatastorageandmanagementcloudGISetc. E-mail:tuzhenfa@126.com 檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪檪 ( 上接第 1338 页 ) LocalTerrainDeformationinPanoramicMaps LI Lin 1 YU Zhonghai 1 LI Zhen 1 XIANG Zhujun 2 (1 SchoolofResourceandEnvironmentalSciencesWuhanUniversity129LuoyuRoadWuhan430079China) (2 Troops61363Xi an710054china) Abstract:Theaccessandnetworktransmissioncostsforgeospatialdataindistributed GIS areimportantfactorsefectingperformance.thecacheisoneoftheefective measuresto solveperformanceproblems.inthispaperwedesignedacachenamed KV-cachebasedon Key-valuestructureto managegeospatialdataeficiently withasimpledatastructure.a KV-cacheconsistsofamemorycachecaled Mem-cacheandafilecachecaledF-cache.Mem- cacheisthefirstlevelcachethecachewrite-back mechanism guaranteesthatdataisbeing writentothef-cacheasynchronously.thef-cacheiscreatedbyasinglelargefilewhilea cacheindexerbasedonb-treeisbuilttospeedupthequery.afreespacemanagerwhichis alsobasedonb-treeisusedtoalocatethedatastorageaddressforf-cache.experiments comparetheread/writeperformanceofkv-cachewiththetraditionalfilecacheinthe100 MbpsLAN.Theexperimentsshowthattheread/writeperformanceofKV-cacheis much beterthanfilecachethusmanaginggeospatialdatamoresecurelyandeficiently. Keywords:distributedGIS;cache;KV-cache;Key-valuestorage;freespacemanager Abstract:Panoramicmapsareusualyusedtoshowthelandscapefrom aspecificdirection and viewing heightin acentralperspective projection.regardless ofthe direction and heightsomelandscapefeatures maybeobscured.inordertoimprovethevisualefectof panoramicmapsamethodforlocalterraindeformationinthehorizontaldirectionbasedon liquefactiondeformationisproposed.this methodforreducinglocalterrainelevationis basedoneyetrackingamethodthatraisesthelocalterrainelevationusingagaussiantrans- form.thismethodreducesocclusionandshrinkagephenomenonenhancingtheartisticvis- ualefectsofapanoramicmap. Keywords:panoramic maps;localterrain distortion;liquefaction deformation;gaussian transform;gazetracking Aboutthefirstauthor:LILinprofessorPhDsupervisor.Heisengagedincartographygeo-ontologyand3Dcadaster. E-mail:lilin@whu.edu.cn