商務(wù)智能解決方案

Sybase/BusinessIntelligenceSYBASE數(shù)據(jù)倉庫/商務(wù)智能解決方案魏健商務(wù)智能咨詢顧問SYBASE軟件(中國)有限公司議程數(shù)據(jù)倉庫解決方案概述數(shù)據(jù)倉庫設(shè)計(jì)工具數(shù)據(jù)倉庫引擎SybaseAdaptiveServerIQMultiplex“數(shù)據(jù)倉庫是在企業(yè)管理和決策中面向主題的,集成的,與時(shí)間相關(guān)的和不可修改的數(shù)據(jù)集合”BillInmon數(shù)據(jù)倉庫定義OLTP系統(tǒng)5-10年過去詳細(xì)數(shù)據(jù)當(dāng)前詳細(xì)數(shù)據(jù)輕度匯總數(shù)據(jù)高度匯總數(shù)據(jù)數(shù)據(jù)集市用戶分析網(wǎng)絡(luò)資源分析數(shù)據(jù)倉庫數(shù)據(jù)倉庫/決策分析系統(tǒng)數(shù)據(jù)倉庫是完全不同的數(shù)據(jù)庫系統(tǒng)RDBMSSybaseSAP/ERPVSAMEXCEL操作(業(yè)務(wù))系統(tǒng)特性事務(wù)處理性能是第一位的支持日常的業(yè)務(wù)事務(wù)驅(qū)動(dòng)面向應(yīng)用數(shù)據(jù)是當(dāng)前的并在不斷變化存儲(chǔ)詳細(xì)數(shù)據(jù)(每一個(gè)事件或事務(wù))針對快速預(yù)定義的事務(wù)優(yōu)化設(shè)計(jì)可預(yù)見的使用模式支持辦事人員或行政人員數(shù)據(jù)倉庫應(yīng)用系統(tǒng)特點(diǎn)支持長遠(yuǎn)的業(yè)務(wù)戰(zhàn)略決策分析驅(qū)動(dòng)面向主題數(shù)據(jù)是歷史的數(shù)據(jù)反映某個(gè)時(shí)間點(diǎn)或一段時(shí)間數(shù)據(jù)是靜態(tài)的，除數(shù)據(jù)刷新外數(shù)據(jù)是匯總的優(yōu)化是針對查詢而不是更新支持管理人員和執(zhí)行主管人員數(shù)據(jù)倉庫解決方案解決從數(shù)據(jù)庫中獲取信息的問題。INFORMATION信息信息INFORMATION什么是數(shù)據(jù)倉庫解決方案？應(yīng)用價(jià)值時(shí)間1.日常報(bào)表2.即席查詢3.分析4.數(shù)據(jù)挖掘?qū)ｎ}應(yīng)用1234數(shù)據(jù)倉庫應(yīng)用類型數(shù)據(jù)倉庫應(yīng)用數(shù)據(jù)倉庫系統(tǒng)體系架構(gòu)RelationalPackageLegacyExternalsourceDataCleanToolSourceDataDataStagingWareHouseAdmin.ToolsEnterpriseDataWarehouseDataExtraction,TransformationandloadDatamartDatamartEnterprise/CentralDataWarehouseRDBMSROLAPRDBMS

DimensionModelingConformeddimension&factIncludingatomic&aggregateArchitectedDatamartsCentralMetadataDataModelingToolEnd-UserToolEnd-UserToolMDBEnd-UserToolEnd-UserToolLocalMetadataLocalMetadata數(shù)據(jù)倉庫/商務(wù)智能應(yīng)用成功的關(guān)鍵?做什么，怎么做？?數(shù)據(jù)倉庫性能Sybase&Partner專專業(yè)業(yè)服服務(wù)務(wù)數(shù)據(jù)據(jù)倉倉庫庫顧顧問問咨咨詢詢IndustryWarehouseStudioSybaseIWS方方法法學(xué)學(xué)ERDesignToolImpactAnalysisMetadataManagementSybaseIndustryWarehouseStudio打打包包的的數(shù)據(jù)據(jù)倉倉庫庫基基礎(chǔ)礎(chǔ)平平臺(tái)臺(tái)概概述述業(yè)務(wù)務(wù)模型型物理理模式式元數(shù)數(shù)據(jù)據(jù)ETL工工具具例子子報(bào)表表算法法ETLToolMetadataExchangeSmartETLMaps(Future)SQLTemplatesCognosBusinessObjectsMicroStrategyBusinessModelsfocusedonKeyIndustryEventsEnterprise-wide,StarSchema-baseddesignIWS產(chǎn)品介介紹TABLETABLETABLETABLETABLEIndustry-specificDataModelsDataWarehouse“OpenRDBMS*”O(jiān)RACLE,IBM,MICROSOFT,NCR,SYBASE,etc.

BIPartnersSampleApplications

AnalyticalCRMSalesAnalysisCustomerProfilingCampaignAnalysisCustomerCareAnalysisLoyaltyAnalysisBusinessPerformanceAnalysisIndustrySpecificSampleDataGeneral-RepresentativeSystemsIntegratorsGuideProjectPlansImplementationProtocole.g.InformaticaETLToolWarehouseArchitectMulti-DimensionalDesignToolSQLSampleReportsWarehouseControlCenterMetaDataManagement客戶構(gòu)成分析營銷活動(dòng)分析客戶興趣分析忠誠度分析銷售分析行業(yè)相關(guān)的經(jīng)營業(yè)績分析收益率分析EVT_TYP_ID=EVT_TYP_IDPRD_ID=PRD_IDENTY_ID=ENTY_IDENTY_ID=EMP_IDGEO_ID=GEO_IDLANGUAGE_ID=LANGUAGE_IDPRODUCT_ID=PRODUCT_IDDEMO_ID=DEMO_IDENTY_ID=V_E_ENTY_IDENTY_ID=ENTY_IDENTY_ID=F_C_ENTY_IDCOR_EVT_TYP_ID=COR_EVT_TYP_IDCOR_RPT_STRC_ID=COR_RPT_STRC_IDENTY_ID=CNTC_RSOL_EMP_IDGEO_ID=GEO_IDFNCL_SCOR_ID=FNCL_SCOR_IDMEASURE_UNIT_ID=MEASURE_UNIT_IDCOR_EVT_TXN_ID=COR_EVT_TXN_IDLANGUAGE_ID=LANGUAGE_IDCOR_EVT_TXN_SEQ_NB=COR_EVT_TXN_SEQ_NBPN_BHVR_SCOR_ID=PN_BHVR_SCOR_IDPRODUCT_ID=PRODUCT_IDDEMO_ID=DEMO_IDENTY_ID=ENTY_IDFNCL_SCOR_ID=FNCL_SCOR_IDMEASURE_UNIT_ID=MEASURE_UNIT_IDDEMO_ID=DEMO_IDPRODUCT_ID=PRODUCT_IDPN_BHVR_SCOR_ID=PN_BHVR_SCOR_IDLANGUAGE_ID=LANGUAGE_IDFNCL_SCORES_ID=FNCL_SCOR_IDMEASURE_UNIT_ID=D_M_MEASURE_UNIT_IDMEASURE_UNIT_ID=MEASURE_UNIT_IDGEO_ID=GEO_IDCOR_RPT_STRC_ID=COR_RPT_STRC_IDEVT_TYP_ID=COR_EVT_TYP_IDENTY_ID=F_C_ENTY_IDGEO_ID=GEO_IDLANGUAGE_ID=LANGUAGE_IDEVT_TYP_ID=EVT_TYP_IDDV_HR_EVT_TYPEEVT_TXN_ID<pk,fk>INTEGEREVT_TYP_ID<fk>INTEGEREVT_TYP_SHRT_NMCHAREVT_TYP_FULL_NMcharEVT_TYP_CAT_SHRT_NCHAREVT_TYP_CAT_FULL_NcharF_HR_EVTV_E_ENTY_ID<fk>INTEGERV_E2_ENTY_ID<fk>INTEGEREVT_DT_PRD_IDINTEGERADMIN<pk,fk>INTEGEREVT_EMP_ID<pk,fk>INTEGEREVT_EMP_DEMO<pk,fk>INTEGEREVT_ADMIN_DEMO<pk,fk>INTEGERCORE_EXT_ID<pk,fk>INTEGERCORE_RPTG_STRUC<pk,fk>INTEGERGEO_ID<pk,fk>INTEGERMU_ID<pk>INTEGERFIN_SCORE_ID<pk,fk>INTEGERLANGUAGE_ID<pk,fk>INTEGERPB_SCORE_ID<pk>INTEGERF_C_ENTY_ID<fk>INTEGERPRODUCT_ID<pk>INTEGERDEMO_ID<pk,fk>INTEGEREMP_ID<pk,fk>INTEGERCDEX_SEQ_NO<pk>INTEGERQTYintegerF_CORE_EVTCOR_EVT_TXN_ID<pk>INTEGERCOR_EVT_TYP_ID<pk,fk>INTEGERD_M_MEASURE_UNIT_ID<fk>INTEGERCOR_RPT_STRC_ID<pk,fk>INTEGERGEO_ID<pk,fk>INTEGERMEASURE_UNIT_ID<pk,fk>INTEGERFNCL_SCOR_ID<pk,fk>INTEGERLANGUAGE_ID<pk,fk>INTEGERPN_BHVR_SCOR_ID<pk,fk>INTEGERPRODUCT_ID<pk,fk>INTEGERDEMO_ID<pk,fk>INTEGERENTY_ID<pk,fk>INTEGERV_E_ENTY_ID<fk>INTEGERCOR_EVT_TXN_SEQ_NB<pk>NUMBERPRD_ID<fk>INTEGERAMOUNTNUMBERD_CORE_EVT_TYPEVT_TYP_ID<pk>INTEGEREVT_TYP_SHRT_NAMVARCHAR(15)EVT_TYP_LONG_NAMVARCHAR(35)EVT_TYP_SUBTYP_NAMVARCHAR(15)D_CORE_RPT_STRCCOR_RPT_STRC_ID<pk>INTEGERHOLDING_COMPANYVARCHAR(35)ORG_TYPEVARCHAR(20)ORG_NAMEVARCHAR(35)REGIONVARCHAR(20)SALES_TEAM_TYPEVARCHAR(15)SALES_TEAMVARCHAR(15)SALES_PERSON_NAMEcharSALES_PERSON_GRADECHARSALES_PERSON_TYPECHARCHNL_CATEGORY1char(18)CHNL_TYPECHARCHNL_SUBCATCHARCHNL_NAMEcharCHNL_CEASED_TRD_DTDATECHNL_ENTY_IDINTEGERCHNL_CITYVARCHAR(20)CHNL_POSTCODEVARCHAR(20)BEGIN_DATE_PRD_IDINTEGEREND_DATE_PRD_IDINTEGERD_GEOGRAPHYGEO_ID<pk>INTEGERALL_ENTRIESCHARPOSTAL_CODECHARVARYING(15)CITYcharPOSTAL_CD_PFXchar(3)HZRD_WTHR_AREACHARHZD_WTHR_TYPECHARDMA_CODECHARSMSA_CODECHARST_PROV_AREACHARTV_REGIONCHARNTL_RADIO_AREACHARLCL_RADIO_AREACHARREGIONCHARCOUNTRYchar(3)CONTINENTY_ABBRchar(3)GEO_SUB_CNTNT_ABBRchar(3)SMRY_EFF_DTINTEGERSMRY_END_DTINTEGERPRISN_ADRS_INDCHARD_MSR_UNITMEASURE_UNIT_ID<pk>INTEGERSHRT_DESCchar(6)LONG_DESCchar(20)D_DEMOGRAPHICSDEMO_ID<pk>INTEGERALL_ENTRIESCHARINCOME_BANDVARCHAR(50)AGE_BANDVARCHAR(50)GNDRCHARMRTL_STATCHARHIGH_VALUE_INDICATCHARACMDTN_CTGRYCHARNBR_IN_HH_BANDVARCHAR(50)CHLD_AT_HOME_BANDVARCHAR(50)SIZE_CLSCHARLEGAL_ORG_TYPECHARNBR_EMP_BANDVARCHAR(50)SECTOR_CLSCHARMAIL_PRMSN_INDCHARTELMKT_PRMSN_INDCHARD_FNCL_SCORFNCL_SCORES_ID<pk>INTEGERINTERNAL_FNCL_SCORVARCHAR(50)EXPERIAN_SCOR_BANDVARCHAR(50)SCOR_N_BANDVARCHAR(50)PRFT_IND_BANDVARCHAR(50)DEBT_INCOME_RATIONUMBERD_LANGUAGELANGUAGE_ID<pk>INTEGERISO_LANG_CODECHARISO_LANG_NAMEcharLANG_GROUPVARCHAR(20)D_PN_BHVR_SCORPN_BHVR_SCOR_ID<pk>INTEGERSCORE1_BANDVARCHAR(20)SCORE_N_BANDVARCHAR(20)D_PRODUCTPRODUCT_ID<pk,fk>INTEGERENTY_ID<fk>INTEGERPRODUCT_LINECHARPRODUCT_GROUPCHARPRODUCT_CODECHARPRODUCT_NAMECHARPD_VARIANT_CODECHARPRODUCT_VARIANTVARCHAR(35)GRP_INDV_INDCHARPD_START_PRD_IDINTEGERPD_END_PRD_IDINTEGERF_SALES_EVENTEVT_TXN_ID<fk>INTEGEREVT_TYP_ID<fk>INTEGERRPT_STRC_ID<fk>INTEGERMEASURE_UNIT_ID<fk>INTEGERFNCL_SCOR_ID<fk>INTEGERPN_BHVR_SCOR_ID<fk>INTEGERENTY_ID<fk>INTEGEREMP_ID<fk>INTEGEREVT_TXN_SEQ_NBR<fk>INTEGERF_CUS_CNTC_EVTV_E_ENTY_ID<fk>INTEGERCUS_CNTC_ID<pk>INTEGERD_C_CTCT_RSOL_ID<fk>INTEGERLGCY_SYS_CUS_CNTCINTEGERCUS_CNTC_REFcharCUS_CNTC_EVT_IDINTEGERF_C_ENTY_ID<fk>INTEGERCUS_STSF_RT_ID<fk>INTEGERCNTC_INIT_DT_IDINTEGERHOUR_ID<fk>INTEGERMINUTE_ID<fk>INTEGERINIT_CNTC_EMP<fk>charCOR_EVT_TXN_ID<fk>INTEGERCOR_EVT_TYP_ID<fk>INTEGERCOR_RPT_STRC_ID<fk>INTEGERGEO_ID<fk>INTEGERMEASURE_UNIT_ID<fk>INTEGERFNCL_SCOR_ID<fk>INTEGERLANGUAGE_ID<fk>INTEGERPN_BHVR_SCOR_ID<fk>INTEGERPRODUCT_ID<fk>INTEGERDEMO_ID<fk>INTEGERCNTC_RSOL_EMP_ID<fk>INTEGERCUS_ID<fk>INTEGERSRSNS_CUS_CO_ID<fk>INTEGERDV_EMPENTY_ID<pk,fk>INTEGERRPT_STRC_IDINTEGERGEO_IDINTEGERADR_IDINTEGEREMP_DEMO_IDINTEGEREMP_NAME_PFXCHAREMP_SNAMEVARCHAR(15)EMP_FNAMEVARCHAR(15)EMP_MNAMEVARCHAR(15)EMP_NAME_SFXCHAREMP_NTL_INS_NBRCHAREMP_HOME_TEL_NBRCHAREMP_PRIM_FAX_NBRCHAREMP_EMAIL_IDINTEGEREMP_DOBDATEEMP_GNDRCHAREMP_MRTL_STATCHAREMP_LIFE_STATCHAREMP_PREF_LANGVARCHAR(20)F_CPGN_CNTC_EVTCCE_ID<pk>INTEGERPROMO_EPSD_ID<pk>INTEGERENTY_ID<pk,fk>INTEGERCNTC_PRD_ID<pk>integerCCH_COUNT<pk>INTEGERCORE__EVT_TYPE_ID<fk>INTEGERCOR_RPTG_STRUCT_ID<fk>INTEGERGEO_ID<fk>INTEGERMU_ID<fk>INTEGERFINANCIAL_SCORE_ID<fk>INTEGERLANGUAGE_ID<fk>INTEGERPB_SCORE_ID<fk>INTEGERPRODUCT_ID<fk>INTEGERDEMO_ID<fk>INTEGEREMP_ID<fk>INTEGERCOR_EVT_TX_SEQ_NO<fk>SMALLINTTRGT_GRPchar(3)CORE_EVENTY_TYPE_IDINTEGERCNTCT_CNTRL_GRP_INCHARCCE_RESULTCHARP_PSYCH_IDINTEGERAFFILIATION_IDintPA_IDINTEGERCC_COMM_EVT_AMTdecimal(10,2)D_TIME_PERIODPRD_ID<pk>INTEGERDT_NAchar(4)DATEDATEDAY_NAMEchar(8)DAY_ABRchar(3)DAY_IN_WEEKSMALLINTDAY_IN_MONTHSMALLINTDAY_IN_YEARSMALLINTWEEK_IN_MONTHSMALLINTWEEK_IN_YEARSMALLINTCLNT_SVC_WK_IN_YRchar(18)MONTH_NAMEchar(10)MONTH_ABRchar(3)MONTH_IN_YEARSMALLINTCALENDAR_QTRchar(6)MONTH_IN_QTRSMALLINTWEEK_IN_QTRSMALLINTDAY_IN_QTRSMALLINTFINANCIAL_QTRchar(6)COMPETITOR_FSCL_YRchar(6)MONTH_IN_FNCL_QTRSMALLINTWEEK_IN_FNCL_QTRSMALLINTDAY_IN_FNCL_QTRSMALLINTSEMI_YEARLYSMALLINTYEAR_NAMEchar(18)YEAR_ABRchar(4)SEASON_NAMEchar(18)SEASON_ABRchar(6)NBR_DAYS_SINCE_90integerHOLIDAY_INDCHARXMAS_HLDY_INDCHAREASTER_HLDY_INDCHARD_CPGN_COM_EVT_TYPEVT_TYP_ID<pk,fk>INTEGERCPGN_COMM_DESCCHAR分析型CRM經(jīng)營業(yè)績管理SybaseIndustryWarehouseStudio分析型應(yīng)用框框架Time資源搜集需求理解業(yè)務(wù)線設(shè)計(jì)模式ETL模板板構(gòu)造分析需求求實(shí)施測試用戶反饋精練測試第二代倉庫典型的數(shù)據(jù)倉倉庫項(xiàng)目從這里開開始SybaseIWS提提供的時(shí)間間上的價(jià)值快速啟動(dòng)數(shù)據(jù)據(jù)倉庫項(xiàng)目搜集需求理解業(yè)務(wù)線設(shè)計(jì)模式ETL模板板構(gòu)造分析查詢詢實(shí)施測試第一代倉庫SybaseIWS從這里開始IWS節(jié)省3到6個(gè)月更多的價(jià)值=更快地訪問信息SybaseIndustryWarehouseStudio

ValueProposition回回顧預(yù)先建立的業(yè)業(yè)務(wù)和物理模模型優(yōu)化了項(xiàng)項(xiàng)目進(jìn)度的安安排和加快了了對數(shù)據(jù)的訪訪問基于經(jīng)過驗(yàn)證證的實(shí)施經(jīng)驗(yàn)驗(yàn)和行業(yè)經(jīng)驗(yàn)驗(yàn)設(shè)計(jì)和方法論論是可擴(kuò)展/可定制的安全企業(yè)范圍數(shù)據(jù)庫獨(dú)立面向行業(yè)集成的模型和和基礎(chǔ)平臺(tái)靈巧節(jié)省資源……一半的投投入節(jié)省時(shí)間……更快的實(shí)實(shí)施節(jié)省資金……降低成本本節(jié)省數(shù)據(jù)倉庫系統(tǒng)統(tǒng)體系架構(gòu)RelationalPackageLegacyExternalsourceDataCleanToolSourceDataDataStagingWareHouseAdmin.ToolsEnterpriseDataWarehouseDataExtraction,TransformationandloadDatamartDatamartEnterprise/CentralDataWarehouseRDBMSROLAPRDBMSDimensionModelingConformeddimension&factIncludingatomic&aggregateArchitectedDatamartsCentralMetadataDataModelingToolEnd-UserToolEnd-UserToolMDBEnd-UserToolEnd-UserToolLocalMetadataLocalMetadataAdaptiveServer??IQMultiplex?是專門為滿足足數(shù)據(jù)倉庫和和商務(wù)智能設(shè)設(shè)計(jì)的高性能能的關(guān)系數(shù)據(jù)據(jù)庫系統(tǒng)。IQMultiplex的主要特點(diǎn)是是：?高可擴(kuò)展性–支持?jǐn)?shù)以千計(jì)計(jì)的并發(fā)用戶戶存取TB級的數(shù)據(jù)。?突破性的速度度–閃電般的查詢詢速度，比傳傳統(tǒng)RDBMS快10~100倍以上。?無限的靈活性性–支持任意類型型的即席查詢詢。?最低的擁有總總成本–高效的數(shù)據(jù)壓壓縮存儲(chǔ)，達(dá)達(dá)到30%~60%；簡單的維護(hù)護(hù)和管理。集成的主要產(chǎn)產(chǎn)品DesignWarehouseArchitectManageSybaseASIQMIntegrateInformaticaEnterpriseConnectReplicationServerPowerMartVisualizeBo、BrioCognosSPSSAdministerWarehouseControlCenterWarehouseControlCentreSybase數(shù)據(jù)倉庫相相關(guān)產(chǎn)品集的的構(gòu)成RelationalPackageLegacyExternalsourceDataCleanToolSourceDataDataStagingWareHouseAdmin.ToolsEnterpriseDataWarehouseDataExtraction,TransformationandloadDatamartDatamartEnterprise/CentralDataWarehouseRDBMSROLAPRDBMSRDBMS,StarSchemaArchitectedDatamartsCentralMetadataDataModelingToolEnd-UserToolEnd-UserToolMDBEnd-UserToolEnd-UserToolLocalMetadataLocalMetadataPowerCenterPowerMartSybaseIQMSybaseIQMBrio/BOPowerMartWarehouseArchitectWCCCognos設(shè)計(jì):成功功的關(guān)鍵數(shù)據(jù)庫的設(shè)計(jì)計(jì)對數(shù)據(jù)倉庫庫系統(tǒng)的整體體性能、裝載載和建立索引的時(shí)時(shí)間以及數(shù)據(jù)據(jù)量的增長等等的影響超過過任何其它方面面。數(shù)據(jù)倉庫設(shè)計(jì)計(jì)在支持分析和和決策的查詢詢環(huán)境中，使使業(yè)務(wù)用戶可可以訪問，理解和和利用數(shù)據(jù)以業(yè)務(wù)用戶理理解和運(yùn)用信信息的方式組組織數(shù)據(jù)可預(yù)見的查詢詢方式基于時(shí)間的匯總的數(shù)據(jù)向下/上的鉆鉆?。―rill-down/drill-up）多維模型設(shè)計(jì)計(jì)傳統(tǒng)的數(shù)據(jù)建建模方法(如如ER模型)可能非非常復(fù)雜且不不易理解按照最終用戶戶的想法定義義信息(以以查詢?yōu)橹行男慕?Star(星星型),Snowflake(雪雪花型)，Constellation(星座座型)，Snowstorm(雪暴暴型)Facts(事實(shí))：可可度量數(shù)據(jù)，，如數(shù)量、、價(jià)格Dimensions(維)：用于于分類Fact的詳細(xì)數(shù)數(shù)據(jù)GroceryTransactionStoreNumberTransactionDateCustomerProductQuantityAmountCustomerCustomerFromDateToDateFirstNameLastNameAddress1Address2Address3CityStateCountryPostalCodeTimeTransactionDateStoreStoreNumberStoreNameCityStateCountryTelephoneProductProductDescriptionCategoryFactTableDimensionTablesDimensionTables多維模型:星星型模式GroceryTransactionStoreNumberTransactionDateCustomerProductQuantityAmountCustomerCustomerFirstNameLastNameAddress1Address2Address3CityStateCountryPostalCodeCustomerCategoryTimeTransactionDateStoreStoreNumberStoreNameCityStateCountryTelephoneRegionProductProductDescriptionCategoryProductCategoryProductCategoryDescriptionRegionRegionDescriptionSalesPeriodPeriodIdentifierSalesPeriodFromDateToDateCustomerCategoryCategoryCustomerCategory為了避免數(shù)據(jù)據(jù)冗余,用用多張表來描描述一個(gè)復(fù)雜雜維在星型模式的的基礎(chǔ)上,構(gòu)構(gòu)造維表的的多層結(jié)構(gòu)多維模型:雪雪花模式GroceryTransactionStoreNumberTransactionDateCustomerProductPurchaseQuantityAmountCustomerCustomerFirstNameLastNameAddress1Address2Address3CityStateCountryPostalCodeCustomerCategoryTimeTransactionDateStoreStoreNumberStoreNameCityStateCountryTelephoneRegionProductProductDescriptionCategoryProductLineSalesPeriodPeriodIdentifierSalesPeriodFromDateToDateCustomerCategoryCategoryCustomerCategoryProductPurchasesProductPurchaseDateSupplyingVendorPurchaseOrderUnitQuantityPurchaseCostVendorVendorVendorNameAddress1Address2Address3CityStateCountryPostalCodeProductInventoryProductWarehouseLocationQuantityOnHandQuantityBackOrderedWarehouseWarehouseAddress1Address2Address3CityStateCountryPostalCode具有多多個(gè)事事實(shí)表表多維模模型:星星座模模式GroceryTransactionStoreNumberTransactionDateCustomerProductPurchaseQuantityAmountCustomerCustomerFirstNameLastNameAddress1Address2Address3CityStateCountryPostalCodeCustomerCategoryTimeTransactionDateStoreStoreNumberStoreNameCityStateCountryTelephoneRegionProductProductDescriptionCategoryProductLineProductCategoryProductCategoryDescriptionRegionRegionDescriptionSalesPeriodPeriodIdentifierSalesPeriodFromDateToDateCustomerCategoryCategoryCustomerCategoryPromotionPeriodPromotionIdPromotionFromDateToDateProductLineProductLineIDDescriptionProductPurchasesProductPurchaseDateSupplyingVendorPurchaseOrderUnitQuantityPurchaseCostVendorVendorVendorNameAddress1Address2Address3CityStateCountryPostalCodeProductInventoryProductWarehouseLocationQuantityOnHandQuantityBackOrderedWarehouseWarehouseAddress1Address2Address3CityStateCountryPostalCode具有多個(gè)事事實(shí)表與多多層維表多維模型:雪暴模模式數(shù)據(jù)模型中中的事實(shí)和和維度事實(shí)和維的的概念對應(yīng)應(yīng)于:數(shù)據(jù)倉庫數(shù)數(shù)據(jù)庫中的的數(shù)據(jù)模型型對象星型模式（（Starschema））DSS/OLAP系統(tǒng)統(tǒng)中的數(shù)據(jù)據(jù)模型對象象多維模型（（Multidimensionalmodel）SalesfactSalesmeasuresTimedimensionAttributesofthe

timedimension星型模式-StarSchemaSalesCubeSalesmeasures(Metrics)TimedimensionAttributesofthe

timedimension多維模型-MultidimensionalModel數(shù)據(jù)倉庫設(shè)設(shè)計(jì)工具WarehouseArchitect為數(shù)據(jù)倉庫庫的設(shè)計(jì)提提供三大功功能：多維建模度量、維、、屬性事實(shí)表，維維表維層次表，，事實(shí)層次次表設(shè)計(jì)向?qū)Ь酆希ˋggregationWizard）分片（PartitioningWizard）逆向工程數(shù)數(shù)據(jù)源優(yōu)化代碼生生成目標(biāo)數(shù)據(jù)倉倉庫引擎（（IQM，，RDBMS)OLAP分析環(huán)境Timeidentifier=TimeidentifierProductidentifier=ProductidentifierCustomeridentifier=CustomeridentifierStoreidentifier=StoreidentifierCustomerCustomeridentifier<pk>doubleCustomernamechar(30)SalesFactProductidentifier<pk,fk>doubleTimeidentifier<pk,fk>doubleCustomeridentifier<pk,fk>doubleStoreidentifier<pk,fk>doubleSalestotalrealProfitsrealStoreStoreidentifier<pk>doubleStorenamechar(50)TimeTimeidentifier<pk>doubleDatetimestampMonthchar(50)QuarterdoubleYeardoubleProductProductidentifier<pk>doubleProductdescriptionchar(80)WarehouseArchitectWarehouseArchitectDataWarehouseorDataMartDatabaseOperationalSourceOLAPEngineInterfaceExternalObjectsDecisionSupport/OLAPModel

(WAMultidimensionalHierarchy)DimensionalAnalysisTransformationRelationaland/orDimensionalAnalysisDataWarehouseModel(WAM)WarehouseArchitect的支持持范圍數(shù)據(jù)倉庫設(shè)設(shè)計(jì)-小結(jié)結(jié)WarehouseArchitect對數(shù)據(jù)倉庫庫設(shè)計(jì)過程程的每一步步都提供支支持：數(shù)據(jù)源中的的元數(shù)據(jù)導(dǎo)導(dǎo)入。設(shè)計(jì)和優(yōu)化化數(shù)據(jù)倉庫庫的數(shù)據(jù)模模型（星型型模式/多多維模型））。與抽取、轉(zhuǎn)轉(zhuǎn)換工具對對接，實(shí)施施數(shù)據(jù)移動(dòng)動(dòng)?；跀?shù)據(jù)倉倉庫模型，，為前端DSS/OLAP工具生成所所需的數(shù)據(jù)據(jù)立方體。。為設(shè)計(jì)過程程的每一步步生成文檔檔和報(bào)告。。數(shù)據(jù)存儲(chǔ)、、管理挑戰(zhàn)數(shù)據(jù)規(guī)模查詢性能裝載速度易于管理存取訪問成功的關(guān)鍵鍵快速，高效效數(shù)據(jù)存儲(chǔ)儲(chǔ)技術(shù)出色的查詢詢性能-特殊的的索引技術(shù)，并行行查詢可伸縮性-GB到TB級易于管理-方便便，靈活，，GUI存取訪問-數(shù)據(jù)據(jù)隨時(shí)可用用數(shù)據(jù)管理解決的方案案通用的關(guān)系系數(shù)據(jù)庫系系統(tǒng)專門的數(shù)據(jù)據(jù)倉庫服務(wù)務(wù)器SybaseIQM專門為數(shù)據(jù)據(jù)倉庫/數(shù)數(shù)據(jù)集市設(shè)設(shè)計(jì)的關(guān)系系型數(shù)據(jù)庫庫專門針對OLAP/DSS而而優(yōu)化的索索引和查詢詢處理技術(shù)術(shù)AdaptiveServerIQM數(shù)據(jù)存儲(chǔ)：：AdaptiveServerIQM垂直存儲(chǔ)技技術(shù)(VerticalPartitioning)無處不索引引(IndexEVERYWHERE)專利的BitWise索引引技術(shù)跨越越Bitmap的限限制多種索引類類型：FP,LF,HNG,HG,CMP,WD低級數(shù)的限限制從100擴(kuò)充到到1000數(shù)據(jù)壓縮(通常達(dá)到到原始數(shù)據(jù)據(jù)的70-75%)預(yù)連接的索索引提供額額外的顯著著提高性能能手段(JoinIndex)支持任意設(shè)設(shè)計(jì)模式星型、雪花花、雪暴、、星座模式式普通關(guān)系模模式支持任意加加載方式文件、內(nèi)部部數(shù)據(jù)、外外部數(shù)據(jù)庫庫直接加載載開放的接口口Index傳統(tǒng)RDBMSRelationalTableTypicalRDBMS數(shù)據(jù)按行存存儲(chǔ)數(shù)據(jù)與索引引分開存放放很少的索引引類型-B-樹普通關(guān)系數(shù)數(shù)據(jù)庫為OLTP系統(tǒng)進(jìn)行優(yōu)優(yōu)化B-treeIndexbestforretrievingonerowatatime計(jì)算“NY”州A類商店的的平均銷售額額當(dāng)表的記錄錄數(shù)從幾萬萬條變?yōu)榍f和上億億條時(shí)，傳統(tǒng)RDBMS技術(shù)術(shù)面對的問問題:表掃描的性性能極端低低下冗余設(shè)計(jì)代代價(jià)高昂、、查詢讀取取的無效字字段過多低級數(shù)類型型數(shù)據(jù)上索索引的失效效普通索引加加載和空間間代價(jià)，造造成不能任任意建造即席查詢的的SQL順順序?qū)π阅苣苡酗@著影影響數(shù)值型比較較和運(yùn)算，，無恰當(dāng)手手段加速處處理傳統(tǒng)RDBMS不適適合數(shù)據(jù)倉倉庫IQM的特特殊存儲(chǔ)方方式-垂直直存儲(chǔ)（按按列存儲(chǔ)））SybaseIQM:數(shù)據(jù)是按列列存儲(chǔ)的，，而不是按按行存儲(chǔ)好處:只存取查詢詢所需的數(shù)數(shù)據(jù)數(shù)據(jù)類型是是一致的，，因而可以以很容易被被壓縮數(shù)據(jù)庫易于于修改和管管理SybaseIQM:只讀完成查查詢所涉涉及到的列列計(jì)算在紐約約的“A””類商店的平均銷售售額好處:

無須使用其他的技術(shù)，SybaseIQM就可以減少I/O超過90%IQM的特特殊存儲(chǔ)方方式-垂直直存儲(chǔ)（按按列存儲(chǔ)））“HowmanyMALESareNOTINSUREDinCALIFORNIA?GenderMMFMM-800Bytes/Row10MROWSStateNY

CA

CTMACA-RDBMSInsuredY

YNYNM Y CAM N CAF Y NYM N CA1243GenderInsuredState++11011101010110MBits10MBitsx3col/816KPage=235I/Os800Bytesx10M16KPage=500,000I/Os基本上只能能使用表掃掃描查詢過程讀讀取了太多多的無效數(shù)數(shù)據(jù)IQMExample:I/O的的明顯減減少IQM的索索引特點(diǎn)索引即是數(shù)數(shù)據(jù)沒有索引和和數(shù)據(jù)的分分別任何一列可可以建立多多個(gè)索引系統(tǒng)保證至至少會(huì)存在在一個(gè)索引引（FP））索引的選擇擇和設(shè)計(jì)主主要基于：：數(shù)據(jù)的級數(shù)數(shù)（離散值值的個(gè)數(shù)））在查詢中的的使用方式式和SQL語語句的順序序無關(guān)索引的種類類FastProjection(FP)數(shù)據(jù)壓縮存存儲(chǔ)根據(jù)數(shù)據(jù)的的特點(diǎn)會(huì)自自動(dòng)使用三三種方式中中的一種LowFast(LF)Bitmap索索引HighNonGroup(HNG)Bit-wise索索引HighGroup(HG)G-Array(包括一個(gè)個(gè)改進(jìn)的B-tree)Compare(CMP)列比較Word(WD)字符串查找找FP索引有有三種內(nèi)部部形態(tài)根據(jù)數(shù)據(jù)級級數(shù)特征，，IQ自動(dòng)選選擇FP中最合適適的一種表表現(xiàn)形式If級數(shù)數(shù)>65536FPindexIf級數(shù)數(shù)<256FFPIndex(Fast-FastProjection)If級數(shù)數(shù)Between256and65536FFFPIndex(Fast-Fast-FastProjection)FP形式1：FPIndex該列的級數(shù)數(shù)超過65536原始數(shù)據(jù)在在磁盤上壓壓縮存儲(chǔ)alphaalphabetagammabetabetaFP形式2：FFPIndex列級數(shù)<256內(nèi)部生成一一個(gè)單字節(jié)節(jié)的lookup表表不僅擁有較較好查詢效效率，同時(shí)時(shí)得到高效效壓縮DataValuesRedBlueGreenRedColorRedBlueGreen12311123332LookupTableDataFP形式3：FFFPIndex列的級數(shù)界界于256和65536之間系統(tǒng)內(nèi)建一一個(gè)雙字節(jié)節(jié)的lookup表表DataValuesRedBlueGreenRedColorRedBlueGreen12311123332LookupTableData1112333