在印度首屈一指的時尚電子商務平臺中塑造數據的未來.pdf

1、2024 Databricks Inc.All rights reserved1SHAPING THE FUTURE SHAPING THE FUTURE OF DATA IN INDIAS OF DATA IN INDIAS PREMIER FASHION EPREMIER FASHION E-COMMERCE PLATFORMCOMMERCE PLATFORMRaghu Ram VengalilRaghu Ram Vengalil2024 Databricks Inc.All rights reserved2PART OF THE LARGEST ePART OF THE LARGEST

2、e-COMMERCE CONGLOMERATECOMMERCE CONGLOMERATEFLIPKART GROUP DELIVERS ACROSS THE EFLIPKART GROUP DELIVERS ACROSS THE E-COMMERCE VALUE CHAINCOMMERCE VALUE CHAINFlipkart Group CompaniesIndias largest online retailer with capabilities to deliver products across all pincodes in India.Indias leading fashio

3、n and lifestyle retailer with distinctive private brands and labels and exclusive international brandsIndias largest supply chain services provider of e-commerce capable of doing half a million shipments/dayWe are one of Indias leading fashion,beauty and lifestyle e-commerce companies,a part of Flip

4、kart-Walmart2024 Databricks Inc.All rights reserved3MYNTRA MYNTRA-INDIAS FASHION&BEAUTY EXPERTINDIAS FASHION&BEAUTY EXPERTStyling IndiaScaleStyling IndiaScale100Mn+App installs8800+Brand Partners2.9 millionStyles19000+Pincodes2024 Databricks Inc.All rights reserved4TECHNOLOGY AT THE CORE OF ORGANISA

5、TIONS TECHNOLOGY AT THE CORE OF ORGANISATIONS FABRICFABRICPlanning and designBrandenablementApp(Storefront)WarehousingDeliverySourcingAssortment and selection planningTrendspottingDesign planningIn-app Brand StoresDeep InsightsPartner UniversityEkart servicesCatalogingOmniEngagement PaymentsSize and

6、 FitEase of discovery Optimal inventory storage Quality checks using computer visionDelivery optimizationRoute optimizationFraud detectionResulting in a delightful customer experience2024 Databricks Inc.All rights reservedMYNTRAS MYFASHIONGPT&MAYAMYNTRAS MYFASHIONGPT&MAYA2024 Databricks Inc.All righ

7、ts reserved6MDP MDP-MYNTRA DATA PLATFORMMYNTRA DATA PLATFORMData Data DemocracyDemocracySelf ServeMultiSelf ServeMulti-tenancytenancyMultiple Processing Multiple Processing EnginesEnginesCost Cost EfficientEfficientData Security Data Security&Complaince&ComplainceData Data QualityQualityAuto Auto He

8、alingHealingFIFO FIFO-Flexible in Flexible in&Flexible Out&Flexible OutData Data DiscoveryDiscoveryGuiding PrinciplesGuiding Principles2024 Databricks Inc.All rights reserved7MDP MDP-MYNTRA DATA PLATFORMMYNTRA DATA PLATFORM1k Data 1k Data pipelinespipelinesData i/o in Data i/o in petabytespetabytes8

9、0 user queriesClick stream events in 80 user queriesClick stream events in BillionsBillions800 unique usersTransactional events in millions10%MoM growth in 800 unique usersTransactional events in millions10%MoM growth in use cases serveduse cases servedDaily statisticsDaily statistics2024 Databricks

10、 Inc.All rights reservedPREVIOUS STATE ARCHITECTUREPREVIOUS STATE ARCHITECTUREClickstream Ingestion Client EventsOLTPManaged TablesServer EventsSDKSDKKBSCKBSCDedupFile FormatsData Lake Azure BlobData Warehouse Azure SynapseFeature storesRedis,AerospikeLow latency aggregatesDruidKafkaLoadersHive Meta

11、storeModel ServingServer Side IngestionServing GatewayBatch pipelinesKafkaDruid IngestionServicesPersonalisationSearchRecommendation NRTUDP(In-house Reporting Platform)2024 Databricks Inc.All rights reserved9WHAT DRIVES US TO SEEK CHANGEWHAT DRIVES US TO SEEK CHANGEPlatform ScalabilityTotal Cost of

12、OwnershipTime to marketCollaboration across data teamsBuild efficient Data Platform 2024 Databricks Inc.All rights reserved10Protocol Standardization,Multi Engine Support&Seamless InteroperabilityUnified Batch&Stream(Ingestion and ETLs)Very Rich Feature Support(Time Travel,Versioning,Snapshotting)Da

13、ta&Storage Optimization,Auto CompactionAggressive Metadata&Stats,Very Low Storage costsLarge OSS Community support,frequent Version ReleasesLow Infra Footprint,Easy Upgrades and Highly Optimized UsageStream(Realtime)Data Processing ModelADVANTAGES WITH DELTA LAKE!ADVANTAGES WITH DELTA LAKE!2024 Data

14、bricks Inc.All rights reservedREDESIGNED ARCHITECTURE WITH REDESIGNED ARCHITECTURE WITH DATABRICKS ECOSYSTEMDATABRICKS ECOSYSTEMClickstream Ingestion Client EventsOLTPServer EventsSDKSDKStreamFeature storesRedis,AerospikeLow latency aggregatesDruidKafkaHive MetastoreModel ServingServer Side Ingestio

15、nServing GatewayBatch pipelinesKafkaDruid IngestionNRTPower BI+DBSQLServicesPersonalisationSearchRecommendation RawCleansedCuratedStreamUDP(In-house Reporting Platform)StreamStream2024 Databricks Inc.All rights reserved12BUSINESS OUTCOMESBUSINESS OUTCOMESDatabricks Data Intelligence Platform has sig

16、nificantly improved our ability to scale data ingestion and reduce ingestion operational costs by over 35%Real time pipelines became 25%faster through reduction in latencies with Spark Streaming pipelines on Databricks helping aid faster user experiences like in-session personalised recommendations25%Performance Improvement35%Cost Reduction2024 Databricks Inc.All rights reserved13THANKYOUTHANKYOUYou can find us at:https:/