凱捷：通過擴展的湖廳架構加快洞察速度.pdf

1、DRIVING SPEED TO INSIGHTS DRIVING SPEED TO INSIGHTS WITH AN EXTENDED WITH AN EXTENDED LAKEHOUSE ARCHITECTURELAKEHOUSE ARCHITECTURERajesh IyerVP,Global Leader of AI&Machine LearningInsights&DataCapgeminiDatabricks2023Context:Need a Data Architecture Reboot?Enterprises are under pressure to monetize d

2、ata.Generative AI and“AI for data”will bring a lot of interest to Py(Spark)based pipelines Problem:Legacy data architecturesEnterprise data architectures need to be a coherent expression of storage and processing strategies purpose-built to hydrate all eight data consumption patterns where unstructu

3、red data is equally important.INTERMODAL CONTAINERStandard,flexible data formatOpen format=enterprise ownershipDISTRIBUTION MODALITYCost-benefit rulesFlexibility to switchINVISIBLE INTELLIGENCEExpose JOINs as true relationshipsEnable full understanding of firmLANGUAGE BARRIERBusiness conceptsPlain w

4、ay to askMAKE EASY TO SHOPEverything for everyonePick a platformANALYSIS PARALYSISMore data=less attentionStart with a pinhole view8Consumption patternsPurpose:Fuel data consumption patternsTodays data architectures are unable to competently accommodate all enterprise data consumption patterns.This

5、vulnerability is exacerbated when unstructured data takes center stage in business use cases.ENTERPRISE REPORTINGStatic reports that sometimes allow users to drill down to a limited extentDASHBOARDS&ALERTSCustomized vehicles for getting access to data that monitors system or business healthSELF-SERV

6、ICE QUERIESReports built from scratch on platforms meant for business analystsQUERY BY CODEReports built from scratch with help from a programmer in an IT groupREAL-TIME MONITORINGReal-time data to effectively manage business threats and opportunitiesAUTONOMOUS ANALYTICSKPI and trend drivers for dec

7、ision makersML TRAININGAllows AI/ML teams to source dataML SERVINGEmbedded into end-user applications,workflows,or experiencesPath:Hydrate Systems of InsightsWe adopt the Lakehouse architecture as the foundation.We extend it to provide a true semantic view and enable autonomous analyticsPROCESSSOURC

8、ESPROCESSING+OPEN FORMAT STORAGE+VIRTUALIZATIONAI&ARAW LANDING LAKEHOUSEBRONZESCHEMADQtimbr BUSINESSLAYERSTABLEREALTIMELAKEHOUSESILVERCONNECTEDDATAADFRAWRESTRICTEDAZURE EXAMPLE:SYANPSE SPARK POOLS OR DATABRICKSOFFLINE FEATURE STOREFEATURE REGISTRYBATCHAD HOCREAL-TIMEAI MLSTREAMTABLETEXTVIDEOIMAGEAUD

9、IOONLINE FEATURESAUTOAutonomous Analytics for Data to DollarsGoing from System and Data Observability to Business Observability-Automagically surface KPI and trend driversInsurance Example:What was the reason my Loss Ratio KPI went from 76%to 78%from Q4/2022 to Q1/2023THANK YOU FOR PARTICIPATING IN THIS SESSION