1、How to use a semantic layer for data and analyticsA Practical GuideTable of Contents3 Introduction:What is a Semantic Layer?4 Where the Semantic Layer Fits5TrendsDrivingtheNeedforaSemanticLayer6TopChallengesaSemanticLayerSolves12IndustryUseCasesforaSemanticLayer17ConsiderationsforChoosingaSemanticLa
2、yer21AFrameworkforEvaluatingSemanticLayerSolutions23AtScaleCloudDataPerformanceBenchmarks25MakingaBusinessCaseforaSemanticLayer26LearnMore3A Practical Guide to Using a Semantic Layer Youmayhaveheardthetermsemanticlayerbefore;itsbeenaroundforsometime.Peopleinventedsemanticlayerstomoldrelationaldataba
3、sesandtheirSQLdialectsintoanapproachableInterfaceforbusinessusers.In1992,BusinessObjectspatentedthetermandformalizedtheirimplementationastheBusinessObjectsUniverseTM.Fromthatpointon,theconceptofmeasureanddimensionsasanabstractionofSQLhasbecomethepreferredlanguageforbusinessusers.Introduction:What is
4、 a Semantic Layer?PrimaryReasonstoAdoptaSemantic Layer Establishasinglesourceofbusiness-vettedmetricsandanalysisdimensions.Ensure data consumers can access governeddatawithtoolsoftheirchoice.Ensureperformantaccesstodataincludingsummarystatistics.Keepcostslowbymanagingcloudinfrastructuresmartlyandkee
5、pingmanagementandadministrationsimpleLetslearnmore!Untilrecently,however,thesemanticlayer was always closely tied to a businessintelligence(BI)platform.AslongasenterprisesremainedwithintheconfinesoftheirBIvendorofchoice,everythingworkedwell.Today,therearemorewaysthanevertoanalyzedata.Longgonearethed
6、ayswheretherewasoneBIplatformtoruleall.Tightlycouplingasemanticlayertoone analytics consumption style no longermakessense.Toexpandonthat,theexplosionofself-serviceBIhasfreedbusinessusersfromrelyingonIT-preparedanalytics,butattheexpenseofdataconsistencyandtrustinanalyticsoutput.Businessdefinitionsand
7、termshavebecomemutable,malleable,andsubjecttointerpretation.Whileitsgreatthatbusinessusersnowhaveself-serviceBItools,theyalsoneedtobeworkingoffofconsistent,high-qualitydata.Thecostofbaddataisenormous;AccordingtoIBM,poordataqualitycoststheU.S.economyastaggering$3.1trillionannually.Luckily,asemanticla
8、yerthatsdecoupled from the point of consumption can help ease these problems with data quality and empowerself-serviceanalytics.Awell-designedsemanticlayercanleadtobetterdata-drivendecisions.Itsacriticalpartofthemodernanalyticsstack.4A Practical Guide to Using a Semantic Layer Where the Semantic Lay
9、er Sits in Your Data StackAsyoucanseeinthisdiagram,the semantic layer sits between the point of analytics consumption and the data warehouseanddatalake.A semantic layer hides the physical complexityfromendusersandprovidesthemwithunderstandablebusinesstermsanduser-friendlydata,insteadofrawSQLanddatab
10、aseschemas.Thislevelofdatavirtualizationmakesdataaccesspossibleforanyanalyticsconsumer.5A Practical Guide to Using a Semantic Layer PowerfulclouddataplatformslikeSnowflake,GoogleBigQuery,AmazonRedshift,Databricks,andMicrosoftAzureSynapsehavebecomethestandardforenterpriseanalyticsstacks.Accordingtoth
11、eAtScale2020BigData&AnalyticsMaturitySurvey,61%ofrespondentscurrentlyoperateclouddataplatforms,and48%planondeployingthemsoon.TrendsDriving theNeedforaSemantic LayerAsthevolumeofdatainthecloudgrows,dataarchitectsareincreasinglybecomingmorecomfortablewithdatalivingindifferentlocationsandindifferentpla
12、tformarchitectures.However,thisgivesrisetoanewchallengeforIT:managingdataaccessandqualityacrossmultiplesilos.Asemanticlayerbecomesacriticalpieceinaclouddataplatformstrategy(orablendedcloudandon-premstrategy).BothdatascientistsandBIusersneedaccesstoclean,understandabledata.Todaysself-servicearchitect
13、uresoftenforceanalyticsconsumerstobecomedatawranglersanddataengineers.Infact,theaveragedatascientistspendsover45%oftheirtimepreparingdataratherthanmodelingit.Askingbusinessusersanddatascientiststodesigntheirownmetricsandanalysisdimensionseverytimetheystartanewprojectisbothamassivewasteoftimeandareci
14、peforchaosandinconsistency.Finally,asemanticlayercanserveasacentralgovernancegatewayacrosstheenterprise,whichiscrucialasthenumberofsilosanddataaccesspointsexplodes.AsemanticlayerservesasasinglepointofaccesssoITcansecuredataandcontrolaccessacrosstheorganization.ThesameBigData&AnalyticsMaturitySurveyr
15、eferencedaboveshowsthatnearly80%ofenterprisesranksecurityandgovernanceascriticaltotheirsuccessinthecloud.6A Practical Guide to Using a Semantic Layer ThetopfivechallengesasemanticlayercansolveThere are common problems that crop up without asemanticlayerfacilitatingdecision-makinginanorganization.Wec
16、angrouptheseproblemsintofiveareas:1.Differentanalyticstoolpreferences2.Lackofdataaccess3.Slowdataintegrationleadingtosiloedsolutions4.InconsistentBIreportsacrossdifferentbusinessunits5.LowdataconfidenceWhatfollowsisadeepdiveintoeachofthesechallengesand an explanation of how a semantic layer can help
17、 solveit.7A Practical Guide to Using a Semantic Layer #1Businessunitshavepreferencesfor different analytics toolsDresnerreportsthatmanenterprisesusethreeormoreBItools,witheachtoolhavingitsownsourceoftruth.Throwinpossibilitiesofinaccuratereportsfrombusinessanalystsormisleadingpredictionsfromdatascien
18、tists,anditseasy to see how multiple tools can lead to multipletruthsandthatsnotagoodthing!Andthepaceofchangeinclouddatawarehousing,BI,andAI/MLhasresultedinaconstantcycleofupgrades,replatforms,and re-factors across different organizations.Fromatime,cost,andbusinessimpactperspective,itshardtokeepupwi
19、ththesechanges.Asemanticlayerneatlysolvesthisproblembyprovidinganalytics-as-a-service(AaaS)toyourbusinessusersanddatascientists.Thisletsyougrantdataaccesstoyourendusersviatheirtoolsofchoicewhilemaintainingdatagovernanceandsemanticconsistency.Largerorganizationshaveatoughertimeimposingasingleanalytic
20、sstandardacrosstheboard.Thiscanbebecauseofthedisruptionofanacquisition,resistancetochange,orfactorsthatlimitmanagementsabilitytoenforceunifiedstandards.8A Practical Guide to Using a Semantic Layer #2Userscomplainabouta lack of access to dataTheresearchsupportsthis,tooGartnerreportsthat87%oforganizat
21、ionshavelowBIandanalyticsmaturity.Youmighthaveabundantdata,butyourdataconsumersstruggletomakesenseofitanditshamperingthespeedatwhichtheycanmakeaccuratedecisions.Asemanticlayereasesthispainbypoweringyourdatamodelwithcrucialcontexttoaiddecision-making.Dataisplentiful,butcoherentdataisanotherstory.Busi
22、nessanalystsanddatascientistscantrelyonjustanydata.Theyneedtounderstandthedatainlogfiles,relationaltables,andotherdatastoresthroughmetadata.Ifthatsmissing,itleadstotimewastedoninterpretationandeveninaccurateresultsthatcanhurtbusinessperformance.9A Practical Guide to Using a Semantic Layer #3 The slo
23、w pace of data integrationdrivesbusinessestoDIYTheresaclearlinkbetweendatadrivendecision-makingandbusinessperformance:MITreportsthatcompaniesinthetopthreespotsintheirindustrywhoapplydata-drivendecisionmakingrealized5%moreproductivityand6%moreprofitthantheirpeers.Thismovetothecloudandriseofbigdatahav
24、epoweredaBIrevolution,leadingtobusinessuserstakingreportinganddataengineeringintotheirownhands.Thisisapositiveshift.Butitalsohasitsdrawbacks,withmanydataplatformsanddatamartsproliferatingeverywhereandmakingdatagovernancedifficult.Suchasituation shows the need for a semantic layer to simplify and str
25、eamline data accessanduse.Businesstodaymovesquickly,andwaitingforacentralizeddatateamtoproducereportsanddashboardsfordifferentdepartmentalusecasesisnotagoodoption.10A Practical Guide to Using a Semantic Layer Ofcourse,havingmultipleBItoolsacrosstheorganizationresultsindifferingresultsforsimilarqueri
26、es.EachBItoolcomeswithitsownmodelinglayer,andallofthemsupportcustomcalculations,soitseasyenoughtocreatewildlydivergentreportsoffofthesamedata.Thatsnotevenaccountingfortablejoinerrors,flawedtime-basedcalculations,orjustsimpleformulamistakes.Thisleadstoacommonconsequence:#4ReportsfromdifferentBItools
27、use similar termsbutshowdifferent results11A Practical Guide to Using a Semantic Layer #5Businessexecsexpressdoubt in the numbersHowever,thisisnttherealityformostbusinessestoday.Manycompaniescannot be sure of the reliability of their data.Thisintroducesdoubtanddelaysindecision-makingasignificantdraw
28、backconsideringthattrustindataisamajorcompetitiveadvantage.Usingonesourceoftruthnaturallyleadstomoretrustinthedata,soifyoufindyourbusinessusersemployingdifferentanalyticstoolstodotheiranalyses,youmaybesufferingfromaconfidencecrisisthatasemanticlayercouldsolve.Thereareseveralapproachestoimplementinga
29、semanticlayerinyourorganizations.Belowisatablewiththeprosandconsforeach:Experianreportsthatsixin10companiesbelievethathigh-qualitydataincreasesbusinessefficiency,44%believeitraisesconsumertrust,43%concludeitenhancescustomersatisfaction,42%believeitdrivesmoreinformeddecision-making,and41%reportthatgo
30、oddatacutscosts.12A Practical Guide to Using a Semantic Layer ACross-IndustryChallengeThepotentialforcreatingvaluewithdata,analytics,andenterpriseAIisacross-industryopportunity.Whileusecasesmayvaryacrossindustryandorganizationanddepartments,theiraresomefundamentalchallengesthatspanallindustries.Inth
31、issection,wellexplorecommonusecasesforasemanticlayerdrawnfromdifferentindustries.13A Practical Guide to Using a Semantic Layer HealthcareManypharmaceuticalandhealthcarecompaniesoperateinhighlycomplexandheavilyregulatedindustries.Asyoumightimagine,theirbusinessesdependondataforsuccess.Someteamschoose
32、tobuildtheirowndataandanalyticsplatformsormakeuseofpre-existingcomponents.Ineitherscenario,asemanticlayerhelpstodemocratizeaccesstodataacrossthecompany.Amongthemanybenefitsofthisapproachisallowinghealthcarecompaniestofocusdataandanalyticseffortsonactivitiesthatimpactprofitandloss.Forapharmaceuticalc
33、ompany,evenasinglepercentagepointofefficiencyimprovementscouldhaveatremendousimpactonmargins.Thegoalistotakeaforward-facing,predictiveapproachtodata,ratherthansimplylookingbackonreportsofwhathasalreadyhappened.Thisapproachalsodramaticallysimplifiestheirdataaccuracy,andreducesreplicationofdataacrossm
34、ultipledatastores.Inaddition,itprovidescommoncontrolsandasharedbacklogsothatbusinessandITteamscandefineworkinbigroomplanningsessionsandpullworkfromacommonbacklogforsprints.Finally,asemanticlayerprovidescrucialsecurityandgovernancecontrols,sothatsensitiveinformationremainsprotected(butmoreonthatlater
35、).14A Practical Guide to Using a Semantic Layer RetailandeCommerce RetailersandeCommerceprovidersrelyontheirdataandinfrastructuretocompete.Withaplethoraofoptionsavailabletoshoppersbothonlineandin-store,theretailerswiththebestdata-drivenstrategiescanprovidehighlytailoredrecommendationsandadapttochang
36、ingcustomerpreferences.Thisagilitystemsfromtheabilityofeveryoneontheteamtobeadataanalyst.Whendealingwithahighvolumeoftrafficandtheresultingmountainofdata,thedatateamstoppriorityisempoweringbusinessuserstoleveragewhicheverdatatoolstheylikebestwhileenablingthemtogetreliable,accurateanswersquickly.Adop
37、tingasemanticlayerfromAtScalehelpsteamsacceleratetimetoinsightfromdata,agnosticoftheirunderlyinginfrastructure.Second,dataproducersneedasetoftechnologiesinordertodotheirjobswell.Thiscouldincludeunderlyingdimensionalmodelsortrainingsetsforamachinelearningmodel.Finally,infrastructurepowerstheactivitie
38、sofbothusergroups(thisincludescomputeenginesandstoragesystemsfordata.)Manylargeretailershaveundertakenatransformationtocloud-basedinfrastructure,whichprovidesaperfecttestcasetouseAtScale.Thegoalistodriveenduseradoptionofcloudtechnologiesthroughtheimplementationofasemanticlayerthatdemocratizedataacce
39、ss.15A Practical Guide to Using a Semantic Layer Consumer PackagedGoodsFordatateamsatconsumerpackagedgoods(CPG)companies,itsanaturalfittoinstitutionalizetheideaofdataasaproduct.Inotherwords,theirteamtreatsdataasapathtounlockingvalueforthebusinessuser.SmartCPGcompaniesleverageconceptsfromengineeringa
40、ndproductmanagementinthesoftwareworldandapplythoseapproachestodata,withimpressiveresults.OnemajorCPGhassuccessfullymanagedtoreducedatasilosandenablebusinessuserstoconsumeitusingasemanticlayer.Theyhavedevelopedalogicalmodelforthebusinessthatservesasasortof“digitaltwin”forthephysicalbusiness.Thisseman
41、tic,logicalmodelmakesitpossibleforbusinessuserstoquerydataandgetanswersusingtermsthattheteamalreadyunderstands.Plus,byleveragingAtScalessemanticlayer,thiscompanycanseparatetheconsumptionofdatafromwherethatdataactuallylivesandhowitisstored.Thisway,datacanliveanywhereandinanyformatwithoutslowingpeople
42、down.Anybusinessusercanaskquestionsandfeelconfidentthattheyarereceivingcorrect,consistentanswers.16A Practical Guide to Using a Semantic Layer FinancialServicesLegacydatastructures,however,cancreateexcessivesilos.Asdatavolumesgrow,businessintelligence(BI),development,anddatabaseengineeringteamsspend
43、significanttimemanagingcachesandmanuallyjoiningdatafromvarioussources.Meanwhile,thebusinesshastobeartheriskassociatedwith penalties due to inaccurate or late reporting.InvestinginasemanticlayercanautomatethemanagementofdataengineeringpreviouslydonebybusyBIteams.Automation can take the manpower out o
44、f datapreparation,byaggregatingrawdatabasedonenduserbehavior.Allofthiscanbedonewhileenhancingexistingsecurityandgovernancecontrols,andmitigatingriskofinaccuratereporting.Asaresult,analystshaveperformantaccessviaasinglesourceoftruth,meetingregulatoryrequirements.Thisrestorestrustinanalyticsandallowse
45、ngineering,BIanddataengineeringteams to spend their time on more productiveactivities.Withasemanticlayer,financialservicesorganizationscansavemillionsintotalcostofownershipfromanalytics,whilesimultaneouslyavoidingtheriskofregulatorypenalties.Duetofederalreportingrequirements,hundredsofanalystsmaynee
46、dtodrilldown across thousands of business calculations to properlysignoffandfilereportsonaregularbasis.17A Practical Guide to Using a Semantic Layer Key Considerations forChoosingaSemantic LayerNowthatyouhaveasenseofhowasemanticlayercansolvecommondatachallengesacrossdifferentindustries,letstalkabout
47、howtogoaboutselectingandimplementingone.Determiningatechnologystrategytoimplementasemanticlayercanbedaunting,butthereareeightkeyconsiderationsto keep in mind as you pick the best approach for your organization.Followingthesegeneralconsiderations,wecomparefouralternativetechnologyapproachesthatorgani
48、zationsmayconsideroutsideofapurpose-builtsemanticlayerplatformlikeAtScale.18A Practical Guide to Using a Semantic Layer #1 NOT TIED TO A SINGLE CONSUMPTION STYLEAsanalyticshavespreadmorewithinorganizations,relyingononeBIorAI/MLplatformtomeeteveryonesneedsisbecominglessrealistic.Also,asemanticlayer t
49、ied to one set of consumption tools isbydesignnot“universal”andinalandscape of many tools and analytics user personas,itscrucialtochooseasemanticlayerdecoupledfromasingleconsumptionstyleoranalyticstool.#2 OFFERS TABULAR AND MULTIDIMENSIONAL VIEWSSemanticlayerscomeintwoflavors:tabularandmultidimensio
50、nal.Thetabular(orrelational)modelbecamepopularinthe70sand80sandreliedon concepts like fact and dimensional tables.Toolsbasedonthismodelweredesignedtomakerelationaldatabasesordatawarehouseseasiertoquery.Multidimensionaldatalayersgoonestepfurtherbydefiningrelationshipsandaggregationrulesandaddingbusin
51、ess-friendlycontextwhilenegatingtheneedforSQL.Itsessentialtochooseasemanticlayertoolthatoffersbothviewstocoverabroaderrangeofusesandconsumptionstyles#3 SUPPORTS DATA PLATFORM VIRTUALIZATIONDatahaslivedinlotsofdifferenthomesovertheyears.Firstitwasthemainframe,thentherelationaldatabase,followedbytheda
52、tawarehouse,theMPPdatabase,thedatalake,andbackagaintothe(thistime,cloud-hosted)datawarehouse.Theseevolutionshavebroughtsignificantchangestohowdataisaccessedandused,andsavvyorganizationshedgeagainstdataobsolescencethroughvirtualization.Virtualizationeliminatesthecostofdatamigrationseverytimeanewtrend
53、gripstheindustry.Asemanticlayervendorshouldofferdatavirtualizationtoabstractawayplatformdifferencesandminimizelock-in.#4 EASY MODEL DEVELOPMENT AND SHARINGRawdataisnear-useless,butaddingadata model to it makes it consumable information.Theidealsemanticlayervendorshouldenableeasyauthoring,sharing,and
54、collaboratingondatamodels.It should also allow the reuse of common objectsandconformeddimensions,theabilitytomodeldatavisually,andacode-basedapproachthatscompatiblewithyourorganizationssoftwaredevelopmentlifecycle.19A Practical Guide to Using a Semantic Layer #5 ABILITY TO EXPRESS DIFFERENT BUSINESS
55、 CONCEPTS AND FUNCTIONSRelationaldataisflexibleandpowerfulbutoftendifficulttoexpresshigh-levelbusinessconstructswith.Theseconstructsincludetime-basedcalculations(e.g.,period-over-period),semi-additivemetrics,ancestor/predecessorfunctions,etc.ExpressingthesecomputationsinSQLischallenging,sochooseasem
56、anticlayerthatsupportsbusinessconstructsandcore analytics requirements around time intelligenceandhierarchicalroll-ups.#6 QUERY PERFORMANCE AND CACHINGQueryperformanceandcachingarecriticalconsiderationsintheselectionprocess.A semantic layer needs consistent and performanttobeofanyusetoitsusers,whoex
57、pectblazinglyfastperformancefromproprietarydatabases.Thisisnteasyconsideringthatmanyoftodaysqueriesoftenincludeheterogeneousdatabasejoinsthatfurthertaxqueryperformance.Toovercomethischallenge,chooseasemanticlayervendorthatincludesacomprehensiveperformancemanagementsystembeyondsimplecachingtechniques
58、.#7 SUPPORT FOR BI AND DATA SCIENCE WORKLOADSTheneedforclean,usabledatadoesntendwithjustthebusinessanalystasreferencedabove,datascientistsspendapproximately45%oftheirtimejustpreppingdataforuse.Acommondatalanguageandbusinesstermsaremorelikely to ensure business analysts and datascientistshavethesamec
59、ontextandproduceconsistentresultsandpredictions.Chooseasemanticlayerthatsupportsvariousworkloads,includingbusinessintelligenceanddatascience.#8 SECURITY&GOVERNANCEBecausethesemanticlayersitsbetweentheorganizationsdataandtheanalyticstoolsthataccessthatdata,theplatformmustintegratewithyourorganization
60、ssecurityinfrastructure.Thiscanhappenintwoways:authenticationandauthorization.First,thesemanticlayermustintegratewithanyexistingsinglesign-oninfrastructuretoauthenticateusers,whetherthroughActiveDirectory,LDAP,OAuth,oranyotherauthenticationplatform.Second,the semantic layer must include the ability
61、tomasksensitivecolumns,limitdatarowsbasedonuseraccessrules,and,crucially,impersonateuserswhenqueryingunderlyingsources.Chooseasemanticlayerthatincorporatesthesetwocriticalsecurityandgovernanceprotocols.20A Practical Guide to Using a Semantic Layer DATA WAREHOUSE/DATA MARTSBUSINESS INTELLIGENCE PLATF
62、ORMSEXAMPLE VENDORSEXAMPLE VENDORSPROSPROSCONSCONSAdatabaseofinformationfromavarietyofdatasourcesTraditionalBIplatformsthatbundledatamodeling,querymanagementandvisualizationSinglesourceoftruthNoextratechnologyneededWidestarrayoftool/queryaccessTightintegrationBusinessuserfriendlyEasy to secureNotfri
63、endlyforbusiness users(tables,columns)Semantic layerspecifictoBItoolonly(notReusable)Slowtointegratenew data sourcesDependence on ITVendor lock in+-+-Snowflake,AmazonRedshift,GoogleBigQuery,AzureSynapseSQLAnalytics.Tableu,PowerBI,IBM,Cognos,SAPBusinessObjects,LookerOLAP PLATFORMSDATA VIRTUALIZATION
64、PLATFORMSEXAMPLE VENDORSEXAMPLE VENDORSPROSPROSCONSCONSAconventionalcube-basedOLAPapproachtoacceleratinganalyticsqueriesbasedonpre-defineddimensionsandmetrics.Platformsthatabstractawaythephysical source and location in a tabularformatAcceleratedqueryperformanceProvidesflexibilityinhow/wheredatais st
65、oredCost efficientGoodintegrationwithMicrosoftBIstack(PowerBIandExcel)Semantic layer can be used acrossavarietyof toolsComplex data pipelines and difficult to make changesSemantic layer is not exposed to otherBIplatformsNotfriendlyforbusiness users(tables,columns)Data is extracted and disconnected f
66、rom EDWDependence on ITData models needto be built beforeaccessingdataQueryperformanceisnotguaranteedand/orneedmanualtuning+-+-MicrosoftSSASorPowerBIPremium,Kyligence,KyvosDenodo,Dremio21A Practical Guide to Using a Semantic Layer A Framework forEvaluatingSemantic Layer SolutionsThefollowingchecklis
67、tincludesdetailedcapabilitiesforyoutoevaluatepotentialsemanticlayersolutions.Theremaybeotherconsiderationsuniquetoyourorganization,butthiscanserveasastartasyouplantrialsandproofofvalueprojects.22A Practical Guide to Using a Semantic Layer TOTALUseCasesFeatureFeature CategoryScore(1-5,5=best)(1-5,5=b
68、est)Weigh(1-5,5=best)Weighted scoreSupportsanalyticalworkloadsSupportsdatascienceworkloadsSupportslegacy,on-premisedatawarehousesSupportson-premiseandclouddatalakesSupportsSaasdatasources(Salesforce,Workday)SupportstoolsthatspeakSQLviaJDBSorODBSSupportstoolsthatspeakMDXorDAXandliveExcelconnectionsSu
69、pportscustomapplicationsviaRESTorPhytoninterfacesSupportsTimeintelligence(periodoverperiod,periodtodate)SupportszeroclientinstallfordataconsumersSupportsMDX,DAX,preandpostquerycalculationsSupportswebbaseddevelopment(versusclientapplication)Supportsaggregationfunctions(SUN,AVG,MAX,MIN)Supportsdialect
70、specificoptimizationsSupportssinglesignonforalldataconsumersSupportsuserimpersonationanddelegatedauthorizationSupportsandrespectsnativedataplatformsecurityconstructsSupportsrowlevelsecurityforusersandgroupsSupportsrowlevelsecurityforusersandgroupsSupportscolumnhidingandmaskingforusersandgroupsSuppor
71、tsmultiple,simultaneouseditorsforvirtualviewdevelopmentSupportsnon-additivemetrics(DistinctCount,First,Last)SupportsreusableobjectsandmodelcomponentsharingSupportsliveExcelpivottablesandExcelCUBEfunctionsSupportsdevelopmentlifestyle(dev/test/prod)SupportsautomatedqueryperformancemanagementConnectivi
72、ty(northbound&southbound)DevelopmentEnvironmentCalculations and Analytical Functions(OLAP)QueryPerformance&CatchingSecurity&Governance23A Practical Guide to Using a Semantic Layer Thefollowingtablessummarizetheseresults.Youcandownloadthefullbenchmarkstudieshere.(embedthislinkhttps:/ Common Cloud Dat
73、a WarehousesModernclouddataplatformshavebeenagamechangerforenterpriseanalyticsprograms.CostefficientcloudstorageandpowerfulSQL-basedqueryenginesmakesitfeasibletomakepetabytesofdataavailableforanalysis.Whendatateamsareproactiveinmakingawiderangeofapplicationdataavailablein“analysis-ready”format,organ
74、izationsarepositionedtodrivetowardafasterrateofdata-driveninsightcreation.HoweverdataavailabilityandSQLaccesstodoesnotautomatically mean data can be deliveredat“speedofthought”orwithintrulyinteractivedatadashboards.Therawscaleofmodernclouddataoftenresultinmulti-secondorlongerquerytimes.Performancela
75、gsleadtoworkaroundsanddata extracts which break semanticconsistency.Thekeytomaintainingasemanticlayerstrategyistodeliverconsistentlyfastperformanceandmanagingcosts.AtScalewasdesignedtodeliverspeedofthoughtperformancewhileeliminatingtheneedtoextractdatalikeconventionalOLAPapproaches.Todemonstratethis
76、,weran20queriesbothwithandwithoutAtScale,usingthestandardTPC-DSbenchmarkv2.11.0fromtheTransactionProcessingCouncil(TPC)forourtests.AtScalesAccelerationStructuresshowedmajorbenefitsinacceleratingqueryperformance-bothrawqueryperformanceandwhenmanagingmultipeconcurrentusers.Wealsopresentmeasuresof simp
77、lification based on the reductioninSQLsyntax.24A Practical Guide to Using a Semantic Layer TESTRedshiftimprovement Factor with AtScaleQueryPerformance12.5xFasterUserConcurrency61x FasterCompute Cost2.6xCheaperComplexity76%LesscomplexSQLquerlesTESTAzure Synapse Analytics SQLimprovement Factor with At
78、ScaleQueryPerformance3x FasterUserConcurrency9xFasterCompute Cost2x CheaperComplexity76%LesscomplexSQLquerlesTESTGoogle BigQueryimprovement Factor with AtScaleQueryPerformance7.7xFasterUserConcurrency20 xFasterCompute Cost10 xCheaperComplexity76%LesscomplexSQLquerlesTESTSnowflakeimprovement Factor w
79、ith AtScaleQueryPerformance4x FasterUserConcurrency14x FasterCompute Cost3.7xCheaperComplexity76%LesscomplexSQLquerles25A Practical Guide to Using a Semantic Layer MakingtheBusinessCasefora Semantic LayerOverthepastseveralyears,organizationshavefocusedtheiranalyticsinvestmentsonembracingmajorcloudda
80、taplatforms.Whilemoderndataplatformssimplifyoperationsandsolvemanyproblems,theydonotaddressthefundamentialproblemsofgettingactionabledatatotheconsumersthatwantit.Buildingabusinesscaseforinvestinginasemanticlayerstrategyshouldfocusonthreeimportantvaluedrivers:1.Increasethevalueofdatabymakingiteasiert
81、ocombinemoredatasets,largedatasets,andwiderwindowsofdatainawaytheyareaccessibletomoredataconsumers.2.Reducethecostofdeliveringanalysis-readydatabyoptimizingcloudspend,simplifyingdataengineeringanddataops,andmakingdataconsumersmoreproductive.3.Increasingthenumberofdataconsumersbymakingiteasiertointer
82、actwithdataassets.Improvingdataliteracyinanorganizationopensuptheflowofdatadriveninsights.26A Practical Guide to Using a Semantic Layer Take the NextStepREQUESTADEMOCONTACTUSLEARNMOREReadtheTechnologyPlatformOverviewforAtScalesSemantic Layers SolutionBuyersGuide:https:/ advicefromfellowdata&analytic
83、sleaders on how to scalesmarterdata-drivendecision-makingStaycurrentonanalyticsstrategieswithshortarticles and webinarsontopicsinanalyticsstrategiesAbout AtScaleAtScaleenablessmarterdecision-makingbyacceleratingtheflowofdata-driveninsights.Thecompanyssemanticlayerplatformsimplifies,accelerates,andex
84、tendsbusinessintelligenceanddatasciencecapabilitiesforenterprisecustomersacrossallindustries.WithAtScale,customersareempoweredtodemocratizedata,implementself-serviceBIandbuildamoreagileanalyticsinfrastructureforbetter,moreimpactfuldecisionmaking.Formoreinformation,andfollowusonLinkedIn,TwitterorFacebook.