工作室天下武功唯快不破saphana實(shí)戰(zhàn)培訓(xùn)

SAPMatchInsightpoweredby“SAP的參與改變了教練、隊(duì)員、球迷和媒體的足球體驗(yàn)。設(shè)想一下：十個(gè)球員用三個(gè)球進(jìn)行有球訓(xùn)練，十分鐘就能產(chǎn)生超過七百萬個(gè)數(shù)據(jù)點(diǎn)，這個(gè)數(shù)字是多么龐大。然而，SAPHANA可以實(shí)時(shí)處理這些數(shù)據(jù)。在SAP的協(xié)助下，我們球隊(duì)可以實(shí)時(shí)分析這些海量數(shù)據(jù)，從而制定培訓(xùn)計(jì)劃并備戰(zhàn)下場(chǎng)比賽。”SAPHANASAPHANASAPHANASAPHANASAPHANASQL實(shí)施SAPHANA

SAPHighPerformanceApplication(基于HANA和SAPUI5

SAP SAP

SAP BatchComputeBatchCompute擎 事件處SAPSAP

儲(chǔ)儲(chǔ)SAPReplicationServer,SAPBusinessObjectsData/

SAPHANA

SAP+

SAPHANA10TB|100SAP數(shù)據(jù)倉庫查詢:300ms-500ms 隨機(jī)分析:SAP數(shù)據(jù)加載DataServices500GB/hr| 數(shù)據(jù)數(shù)據(jù)

SAPHANAStudio基于?2014SAPAG.Allrights SAPHANA

SAPHANA1

appsfor(&分析

on

on

SAPHANABOBJSAPSAPHANABOBJSAPHANARDSHANAFinance&BIon基于AmazonWebServicesbyHANAplanning&OneononSAPHANA作為OLAPSAPHANA1

ETL

行存

SAPSAPHANA1

內(nèi)存容量：128GB~50TB~單機(jī)從128GB~單機(jī)從128GB~4TB100YOYSAPHANA析，9省集中MSS系統(tǒng)-BWonHANA900500SAPHANABIONHANA2011-6?2014SAPAG.Allrights SAPHANA&SAPHANA SAPHANASAPHANASAPHANASAPHANASAPHANASQL實(shí)施SAPHANAClients(planned,BI4Clients(planned,BI4MSBI4ERPSAPBusinessObjectsSBOforSBODiskDataLogPersistencePageRowCalcSQLSQLSessionOtherSource3rd?2014SAPAG.Allrights TheIMC MS Clients(planned,eg SessionSession

I SBOPRequestProcessing/ExecutionRequestProcessing/ExecutionCalcSQLSQLERP PersistencePagePersistencePage

RowRowfor ?2014SAPAG.Allrights LoadingDataintoSAP

MS

Clients

BI4BI4BI4RequestRequestProcessing/ExecutionControlSQLParser SQL Calc

SessionSessionBusinessObjectsBusinessObjects

for

RowPersistencePersistencePageDiskDataDiskDataLog ?2014SAPAG.Allrights Data

MS

BI4BI4BI4 SessionSessionERP BusinessBusinessObjects

SBOSBOfor

RowRowPagePersistencePage

DiskDataDiskDataLog3rd?2014SAPAG.Allrights MSClientsMSClients(planned,BI4BI4ERPDiskDataLogPersistencePageSessionSQLParser SQL CalcRow SBOSBOforBusinessObjects3rd?2014SAPAG.Allrights

MS

Clients(planned,

BI4BI4BI4 SessionSessionCalcSQLSQLBusinessObjects SBOSBOforOtherSource3rd?2014SAPAG.Allrights StorageSeparation(Main& EnableshighcompressionandhighwriteOnlyindeltastoragebecausewriteperformanceatthesame Theupdateisperformedbyinsertingaentryintothedelta

Main

DataCompressioninMainCompressionbycreating

ReadAlwayshavetoreadfrommain&deltastoragesandandapplyingfurthercompressionSpeedDataloadintoCPUEqualitycheck→

theEngineusesmultiversionconcurrencycontrol(MVCC)toensureconsistentreadoperations.?2014SAPAG.Allrights?2014SAPAG.Allrightsdeltamerge

DeltaMergeSeenextColumnStoreDeltaManagementDeltaMergeTomovechangesindeltastorageintothecompressedandreadoptimizedmainHappensEvenduringmergeoperationthecolumnartablewillbestillavailableforreadandwriteTofulfilthisrequirement,aseconddeltaandmainstorageareusedBefore AfterMergeMerge

?2014SAPAG.Allrights PurposeandWhyDoesAnIn-memoryDatabaseNeedAPersistenceMainMemoryisvolatile.WhathappensPower→Dataneedstobestoredinanon-volatileBackupandSAPin-memorycomputingengineoffersonepersistencelayerwhichisusedbyrowstoreandcolumnstoreRegular→fullpersistedimageofDBattimeofLogscapturingallDBtransactionssincelastsavepoint(redologsandundologs→restoreDBfromlatestsavepointAbilitytocreate→usedforSystemRestartandPopulationofIn-memoryActionsDuringSystemLastsavepointmustberestoredUndologsmustbereadforuncommittedtransactionssavedwithlastRedologsforcommittedtransactionssincelastsavepointCompletecontentofrowstoreisloadedintomemoryColumnstoretablesmaybemarkedforpreloadornotOnlytablesmarkedforpreloadareloadedintomemoryduringIftableismarkedforloadingondemand,therestoreprocedureisinvokedonfirstLookandInformationModelerDatabaseChoicetopublishandconsumeat4levelsofDataDataSourceschemas(metadata)–massandselectiveDataProvisioningforSAPBusinessApplications(bothinitialloadandreplication)AnalyticPrivileges/SecurityInformationModelerAttributes–descriptiveMeasures–datathatcanbequantifiedandAttributeViews–i.e.AnalyticViews–i.e.CalculationViews–similartovirtualproviderwithservicesconceptinLeveled–basedonmultipleParent-childAnalyticPrivilege–securityNavigatorView-HANAInstanceHANAServerNameandInstanceNumberUserDatabaseSchemaContent:ColumnViews,NavigatorView-ContentisorganizedinPackagesAttributeViews,AnalyticViews,CalculationViews,AnalyticPrivilegesorganisedinfolders?2014SAPAG.All?2014SAPAG.Allrights ??2014SAPAG.Allrights SAPHANASAPHANASAPHANASAPHANASAPHANASQL實(shí)施SAPHANA

??2014SAPAGoranSAPaffiliatecompany.Allrights SAPHANA123456…SAP

2345

SAPHANA數(shù)據(jù)表分區(qū)，是指在行方向上對(duì)一張完整的數(shù)據(jù)表按照一定的邏輯循序進(jìn)行劃分，分區(qū)的子表作為一個(gè)整體，對(duì)于大多數(shù)SQL查詢和數(shù)據(jù)庫操作語句(DML在大型多節(jié)點(diǎn)SAPHANA系統(tǒng)上，利用數(shù)據(jù)表的分區(qū)操作可以對(duì)巨大體積數(shù)據(jù)表的管理帶來便利，均勻分布系統(tǒng)負(fù)載，優(yōu)化應(yīng)用運(yùn)行性能…等等?2014SAPAGoranSAPaffiliatecompany.Allrights 并不僅僅是多節(jié)點(diǎn)SAPHANA數(shù)據(jù)庫可以利用數(shù)據(jù)表分區(qū)，對(duì)單節(jié)點(diǎn)SAP123 >256…

<2<2<2<2456<26SAP?2014SAPAGoranSAPaffiliatecompany.Allrights 進(jìn)行了表分區(qū)之后，子表共享metadata但是卻擁有獨(dú)立的mainindexdelta少deltamerge操作對(duì)整個(gè)系統(tǒng)資源的占用情況INSERTINTOTABLE… NodeNodeNodeNodeDelta 進(jìn)行了表分區(qū)之后，子表共享metadata,但是卻擁有獨(dú)立的mainindex,deltadelta日志。因此當(dāng)數(shù)據(jù)更新只是在個(gè)別子表上進(jìn)行的時(shí)候，適當(dāng)?shù)膶?duì)表進(jìn)行分區(qū)，將減少deltaSELECTSELECTFROMWHERE國(guó)家=XXXNodeNodeNode時(shí)候，所有存儲(chǔ)該表子表的服務(wù)器會(huì)并行執(zhí)行Query，利用Scaleout的SAPSELECTSELECTFROMWHERE月份=1月NodeNode2NodeINSERTINTOTABLES… NodeNodeNodeNode SAPHANASAPHANA的數(shù)據(jù)表分區(qū)方案強(qiáng)烈的依賴如：系統(tǒng)負(fù)荷，數(shù)據(jù)表結(jié)構(gòu)，SQL查詢方式等等，利用HANA提供的系統(tǒng)數(shù)據(jù)視圖(如:M_CS_TABLES)定期檢查并預(yù)估數(shù)據(jù)表的體積，SAP建議數(shù)據(jù)表的分區(qū)至少在數(shù)單獨(dú)從系統(tǒng)優(yōu)化角度進(jìn)行表的分區(qū)。例如利用數(shù)據(jù)表的分區(qū)減少DeltaMerge對(duì)系統(tǒng)資源占用，但是可能會(huì)大大降低應(yīng)對(duì)SAPNetWeaverBusinessWarehouseonHANA的表進(jìn)行分區(qū)。BWonHANA有自動(dòng)表分區(qū)的機(jī)制.除非SAP支持的建議，CREATECOLUMNTABLEmytab(aINT,bINT,cINT,PRIMARYKEY(a,b))PARTITIONBYHASH(a,b)PARTITIONS4;COLUMNmytabINT,bINT,INT,PRIMARYPARTITIONHASHb)ALTERTABLEmytab_hashPARTITIONBYHASH(v_jldw)PARTITIONSSELECT*FROMmytab_HASHWHEREHANA系統(tǒng)會(huì)根據(jù)SQLQuery的Where語句里包含的過濾條件，對(duì)CREATECOLUMNTABLEmytab(aINT,bINT,cINT)PARTITIONBYROUNDROBINPARTITIONS4RangeRange分區(qū)操作要求對(duì)表統(tǒng)會(huì)報(bào)錯(cuò)。因而需要預(yù)定義一個(gè)被稱為’RestPartition’的分區(qū)子表來接受這些記錄 CREATECOLUMNTABLEmytab(aINT,bINT,cINT,PRIMARYKEY(a,b))PARTITIONBYRANGE(a)(PARTITION1<=VALUES<PARTITION5<=VALUES<20,PARTITIONVALUE=44,PARTITIONOTHERS)繁DeltaMergeHashRangeHashRangeRound-RobinRangeHashHashHashRange分 HashRange分區(qū)是最典型的一種多層分區(qū)方式，如：在第一層分區(qū)利用Hash分區(qū)的時(shí)間分區(qū)原則降低每個(gè)節(jié)點(diǎn)的DeltaMerge帶來的資源消耗，如下圖所示CREATECOLUMNTABLEmytab(aINT,bINT,cVARCHAR(10),PRIMARYKEY(a,b))PARTITIONBYHASH(a,b)PARTITIONS3,RANGE(c)(PARTITIONVALUE=‘JAN’,PARTITIONVALUE=‘FEB’,PARTITIONVALUE=‘MAR’,PARTITIONOTHERS)Round-RobinRange分區(qū)及HashHash分 Round-RobinRangeCREATECOLUMNTABLEmytab(aINT,bINT,cINT)PARTITIONBYROUNDROBINPARTITIONS4,RANGE(c)(PARTITION1<=VALUES<5,PARTITION5<=VALUES<HashHashCREATECOLUMNTABLEmytab(aINT,bINT,cINT,PRIMARYKEY(a,b))PARTITIONBYHASH(a,b)PARTITIONS4,HASH(c)PARTITIONS7對(duì)于表的分區(qū)的監(jiān)控，SAPHANA一共提供了如下四張系統(tǒng)視圖可以查詢到目標(biāo)表的分區(qū)方案明細(xì)SELECTSCHEMA_NAME,TABLE_NAME,PARTITION_SPEC,TABLE_OID,IS_COLUMN_TABLEFROMTABLESWHEREPARTITION_SPECISNOTNULLSELECTHOST,SCHEMA_NAME,TABLE_NAME,PART_ID,MEMORY_SIZE_IN_MAIN,MEMORY_SIZE_IN_DELTA,RECORD_COUNTFROMM_CS_TABLESORDERBYHOST,SCHEMA_NAME,TABLE_NAMEDESCSELECT*FROMM_TABLESWHERESELECTSCHEMA_NAME,TABLE_NAME,PARTITION,SUBPARTITION,RANGEFROMM_CS_PARTITIONSORDERBYSCHEMA_NAMEASC,TABLE_NAMEASC,PARTITIONASC,SUBPARTITIONASCClientSideStatement例如:從Hash(X)3到Hash(X)2例如從Hash(X)3到Hash(Y)拆分具體場(chǎng)景從分區(qū)子表數(shù)n到分區(qū)子表數(shù)m,且例如Hash(X)3到Hash(XALTERTABLEmytabPARTITIONBYHASH(a)PARTITIONSALTERTABLEmytabMERGE對(duì)數(shù)據(jù)表記錄擁有長(zhǎng)時(shí)間的鎖（僅SELECT操作是允許的在操作開始之前，會(huì)先進(jìn)行DeltaMergeHANA的StatisticsServer會(huì)保持對(duì)數(shù)據(jù)增長(zhǎng)和表的體積的監(jiān)控，參照分區(qū)的拆分與合并一節(jié)，進(jìn)行分區(qū)表的合并與再分區(qū)可能需要對(duì)分區(qū)子表進(jìn)行移動(dòng)。割子表，對(duì)多層分區(qū)需要將整個(gè)分區(qū)組進(jìn)行移動(dòng)，建議在進(jìn)行分區(qū)子表的移動(dòng)之前，查看M_C_AITIOSABLE_LOTION了解當(dāng)前分區(qū)子表或分區(qū)組所在位置信息ALTERTABLEmytabMOVEPARTITION2TOALTERTABLEmytabMOVETOCREATECOLUMNTABLEmytab(aINT,bINT)PARTITIONBYHASH(a)PARTITIONS2,RANGE(b)(PARTITION1<=values<5,PARTITIONOTHERS)ALTERTABLEmytabMOVEPARTITION2TOCREATEROWTABLEmytab(IINTPRIMARYKEY)ATALTERTABLEmytab5ADDREPLICAATLOCATIONClientSideStatement境，SQLQuery會(huì)被重新定向和發(fā)QueryPlan和連接線程的運(yùn)行信息數(shù)據(jù)表所在節(jié)點(diǎn)發(fā)送SQLQuery執(zhí)行請(qǐng)求，這也被稱為ClientSideStatementRoutingClientSideStatementRouting在多節(jié)點(diǎn)系統(tǒng)上，表的分區(qū)和移動(dòng)都會(huì)改變，因此系統(tǒng)會(huì)將queryplan重新進(jìn)行編查詢的優(yōu)化，需要設(shè)置HANAServer段和Client端的相關(guān)參數(shù)***在對(duì)表進(jìn)行分區(qū)針對(duì)應(yīng)用進(jìn)行優(yōu)化之前，建議檢查ClientSideStatementRouting機(jī)制是否正常，可以通過系統(tǒng)視圖M_CONNECTIONS查詢客戶端和HANAServer的連接建立信息ClientSideStatementRoutingClientlibrary(SQLDBC,

EnabledNoclientdistributionis

Connectiondistributionis

Both Statementroutingis Statementroutingis Enabled Noclient Noclientdistributionis

Statementrouting Statementroutingis

ConnectiondistributionisBothClientSideStatementStringUSERNAME="username";StringPASSWORD=Propertiesprops=newProperties();props.put("user",USERNAME);…sqlConn=DriverManager.getConnection(DB_URL,propsmode參數(shù)僅提供測(cè)試，不建議在SAPHANASAPHANASAPHANASAPHANASAPHANASQL實(shí)施SAPHANARowStorevs.ColumnStoreWhentoUseWhichStoreModelingOnlyPossibleForColumnThisanswersthefrequentlyasked"WhereshouldIputatable–rowstoreorcolumnInformationModeleronlyworkswithcolumnReplicationservercreatestablesincolumnstoreperDataServicescreatestablesincolumnstoreperSQLtocreatecolumntable:"CREATECOLUMNTABLEStorecanbechangedwith"ALTERTABLESystemTablesAreCreatedWhereTheyFitAdministrativetablesinrowSchemaSYS→caches,administrativetablesofAdministrativetablesincolumnSchema_SYS_BI→metadataofcreatedviews+masterdataforSchema_SYS_BIC→somegeneratedtablesforSchema_SYS_REPO→e.g.listsofactive/modifiedversionsofModelingProcessPhysicaltablesarecreateddynamically(1:1ofsourcesystem

arecreatedAttributeAnalytic

ConsumewithBICS,SQL,Physicaltablesareloadedwith

ColumnviewsFrameworkforModelingwithin Tables/AttrTables/Attr Base Only

not

SchemaorCalcViewsCalcViews

TryCalc

notAttributeAttributeWhatisanAttributeAttributesaddcontexttoAttributesaremodeledusingAttributeCanberegardedasMasterDataCanbelinkedtofacttablesinAnalyticalAmeasuree.g.weightcanbedefinedasanTableJoinsandJoinfullOuter,textTableLanguageAnalyticalAnAnalyticalViewcanberegardedasaAnalyticalViewsdoesnotstoreanydata.ThedataisstoredincolumnstoreortableviewbasedontheAnalyticalViewStructure.AttributeandCancreateAttributeMusthaveatleastoneMusthaveatleastoneCancreateCalculatedCanrenameAttributeandMeasuresonthepropertytabAnalyticalView:DataTherearethreemainviewsonecanselectfromwhenpreviewingRawData–tableformatofDistinctValues–graphicalandtextformatidentifyinguniqueAnalysis–selectfields(attributesandmeasures)todisplayingraphicalNoSQLcodingUnion,Join,ProjectionnodesJoinColumnTables(Analytical SQLorSQLScriptrequiredtocreateScriptbasedCalculationWriteSQLSelectstatementsagainstexistingrawtablesorColumnStoresDefineoutputstructure,activationcreatescolumnstorebasedonScriptCalcEngineforTheeasiestwaytothinkofCalculationModelsistoseethemasdataflowgraphs,wherethemodelercandefinedatasourcesasinputsanddifferentoperations(join,aggregation,projection,…)ontopofthemfordataTheCalculationEnginewillbreakupamodel,forexamplesomeSQLScript,intooperationsthatcanbeprocessedinparallel(rulebasedmodeloptimizer).Thentheseoperationswillbepassedtothedatabaseoptimizerwhichwilldeterminethebestplanforaccessingroworcolumnstores(algebraictransformationsandcostbasedoptimizationsbasedondatabasestatistics).SQLScript/R/BFL(BusinessFunctionThreewaystoimplementin-memorydataminingandstatisticalSQLScriptisasetofSQLextensionswhichallowdeveloperstopushdata-intensivelogicintothedatabaseinordertoavoidmassivedatacopiestotheapplicationserverandtoleveragesophisticatedparallelexecutionstrategiesoftheRThroughtheRintegrationsolution,developerscanleverageopensourceR’s3000+externalpackagestoperformwide-rangedataminingandstatisticalBFL(BusinessFunctionBFListhecalculationlibraryfortheapplicationsbuiltontopoftheSAPHANAdatabase.ThebusinessfunctionsarewritteninC++andexecutedindatabasecalculationengine.BFLhasaroadmapfordataminingandstatisticalalgorithms.ThesetofSQLextensionsfortheSAPHANAdatabasewhichallowdeveloperstopushdataintensivelogicintothedatabaseiscalledSQLScript.TheseextensionsarekeystoavoidingmassivedatacopiestotheapplicationserverandtoleveragesophisticatedparallelexecutionstrategiesoftheSQLScriptV2supportsstoredprocedures,whichprovidesenhancedcontrolflowcapabilitiesandispositionedtobemoresuitableforpushingcomplexpartsofapplicationlogictothedatabase.Itcanmeetsomesimplerequirementforreporting,likejoin,aggregation,etc.Whenitcomestodataminingandstatisticanalysis,SQLScriptisnotsuitableforimplementingcomplexalgorithms.SAPHANASAPHANASAPHANASAPHANASAPHANASQL實(shí)施SAPHANA地利用SAPHANA的復(fù)雜并行執(zhí)行機(jī)制,列存儲(chǔ),查詢優(yōu)化等特性.,SQLScript不適合用于復(fù)雜的算法實(shí)現(xiàn),.SQLScript函數(shù)可以返回多個(gè)結(jié)果集但是SQL加了可讀性.標(biāo)準(zhǔn)SQL雖然支持建立SQL視圖,但是SQL視圖不支持輸入?yún)?shù)通過SQLScript可以用本地變量來保存臨時(shí)結(jié)果而無需定義變量類型,SQLScript是SQL的擴(kuò)展→Insert,Update,Delete,DDL,IF/Else,游標(biāo)loopsbig_pub_ids=SELECTpublisherFROMbooks --QueryQ1GROUPBYpublisherHAVINGCOUNT(isbn)>big_pub_books=SELECTtitle,price --QueryQ2FROM:big_pub_ids,publishers,booksWHEREpub_id=pidANDpub_id=publisherANDcrcy=:currency;output_pubs=SELECTSUM(price) --QueryQ3FROM:big_pub_booksGROUPBYpublisher;output_year=SELECTSUM(price) --QueryQ4FROM:big_pub_booksGROUPBYyear;

BEGIN =BEGIN =*–p_out2*from–END;p_out3*fromSELECTcolumnout=SELECTA,B,CfromSELECTattributeout=SELECTA,B,Cfromout=SELECTA,B,C,SUM(D)from"ANALYTIC_VIEW"GROUPBYA,B,CWHEREout=SELECTA,B,C,SUM(D)from"ANALYTIC_VIEW"WHEREB='value'ANDC=col_tab=CE_COLUMN_TABLE("COLUMN_TABLE");out=CE_PROJECTION(col_tab,[A,B,C],'"B"=''value''AND"C"=GROUPout=SELECTA,B,C,SUM(D)FROM"COLUMN_TABLE"GROUPBYA,B,CCE_AGGREGATION((col_tab,SUM(D),[A,B,C]);INNERout=SELECTA,B,Y,SUM(D)from"COLTAB1"out=CE_JOIN("COLTAB1","COLTAB2",[KEY1,KEY2],[A,B,Y,LEFTOUTERout=SELECTA,B,Y,SUM(D)from"COLTAB1"LEFTOUTERJOIN"COLTAB2"WHERE"COLTAB1"."KEY1"="COLTAB2"."KEY1"AND"COLTAB1"."KEY2"="COLTAB2"."KEY2"KEY2],[A,B,Y,D])SQLout=SELECTA,B,C,SUBSTRING(D,2,5)FROMCE_PROJECTION(:proj_tab,["A","B","C",col_tab1=SELECTA,B,C,DFROM"COLUMN_TABLE1";col_tab1=SELECTA,B,C,DFROM"COLUMN_TABLE1";col_tab2=SELECTA,B,C,DFROM"COLUMN_TABLE2";out=SELECT*:col_tab1UNIONALLSELECT*FROMcol_tab1=管理窗口→追蹤配置（選項(xiàng)卡）→全局(ERROR),警告(WARNING),信息(INFO)或SQL追蹤管理窗口→追蹤配置（選項(xiàng)卡管理窗口→追蹤配置（選項(xiàng)卡）→句法:ylt_tab=select*fromsqlscript_trace;select*from<temp_table>;calltruncate_sqlscript_trace;編譯調(diào)試信息ALTERPROCEDUREgetOutputRECOMPILEWITH理解數(shù)據(jù)流結(jié)構(gòu)和變量的重命名select*fromSYS.PROCEDURE_DATAFLOWSasSYS.PROCEDURE_MAPPINGaswheredf.dataflow_name=cedure_name=‘GETOUTPUT‘orderCALLgetOutput(0,'EUR',NULL,NULL)WITHOVERVIEWINselect*fromselect*from使用工具，比如JSONViewer避免連接(JOIN)輸入數(shù)據(jù)源中不帶WHERE條件的視圖,|constant’skey|intvalue|doublevalue|:SAPHANASAPHANASAPHANASAPHANASAPHANASQL實(shí)施SAPHANASAPHANA設(shè)備方法論（安裝和升級(jí)）永久化存儲(chǔ)備份及恢復(fù)（系統(tǒng)備份容災(zāi)能力期隱患/SAPHANASAPSAP

SAPSolutionmanagerSAPHANA作為設(shè)備提供了E2E的SAPHANA監(jiān)控和管理?2014SAPAG.?2014SAPAG.Allrights SAPHANA技術(shù)合作伙 SAPHANA硬件合作伙SAPHANASAPHANAPlatformSAPHANAEnterpriseSAPHANAfor2x10WestmereEX2Intel處理器)128GB160GBPCIe-FlashSSD1TBSASSSD3x1GBn/wor1x10GBn/w總線),S2x10WestmereEX24Intel處理器256GB320GBPCIe-FlashSSD1TBSASSSD3x1GBn/wor1x10GBn/w總線),M4x10WestmereEX48Intel處理器512GB640GBPCIe-FlashSSD2TBSASSSD3x1GBn/wor1x10GBn/w總線),L8x10WestmereEX8Intel處理器1TB1.2TBPCIe-FlashSSD4TBSASSSD3x1GBn/wor1x10GBn/w總線),/pam(“HANA?2014SAPAG.Allrights SAPHANA“on-premise”

“SaaS設(shè)備運(yùn)營(yíng)的概念基礎(chǔ)設(shè)施SAP

?2014SAPAG.?2014SAPAG.AllrightsSAPHANABI連接 SAPHANASPSAPSAPHANASAPHANAStudioSAPHANAServiceSMP(SPSAPHANAAllremoteconnectionsaredoneviasecurewebIT(ITSAPHANASAPSAPHANA設(shè)備內(nèi)容的技術(shù)操作手冊(cè)(/hana本手冊(cè)會(huì)衍生出操作SAPHANA必須的定期職責(zé)SAPHANA設(shè)備方法論（安裝和升級(jí)）容災(zāi)能力監(jiān)控和管理SAPHANA SAPHANA（這里指redo日志）通過記錄數(shù)據(jù)庫事

1

323 SAPHANA設(shè)備方法論（安裝和升級(jí)）備份及恢復(fù)（系統(tǒng)備份容災(zāi)能力SAPHANA服務(wù) （外部備份

（外部備份

SAPHANAStudioSAPHANAStudio務(wù)務(wù)BackintforSymantecIBMHP(Data??2014SAPAG.Allrights 這是計(jì)劃的當(dāng)前狀態(tài)，SAP隨時(shí)可能修改 從SAPHANASPS4SID數(shù)SAPHANA設(shè)備方法論（安裝和升級(jí)）備份及恢復(fù)（系統(tǒng)備份監(jiān)控和管理HANA

例如向外擴(kuò)展（ScaleM統(tǒng)計(jì)服務(wù)（僅在一臺(tái)激活的節(jié)點(diǎn)上XSAPHANASAPHANA

SAPHANA…管理HANA軟件升級(jí)

SAPHANAStudio 數(shù)據(jù) 數(shù)據(jù) 數(shù)據(jù) 數(shù)據(jù) 數(shù)據(jù) 點(diǎn)HP,Fujitsu, 點(diǎn)不同的案例（HANASAPBW）對(duì)最小安裝可能有不同的需求和建議。（比如BW在SAPnote1637145attachedPDF中有不同的定義）

SAPHANA設(shè)備方法論（安裝和升級(jí)）??2014SAPAG.Allrights 這是計(jì)劃的當(dāng)前狀態(tài)，SAP隨時(shí)可能修改 SAPHANASAPHA