当前位置:首页 > 报告详情

A204_Nataranjan.pdf

上传人: 拾起 编号:1235378 2026-05-04 30页 8.18MB

1、D B TA D ATA S U M M I T 2 0 2 6 B R E A K O U TSimplifying DataInteroperabilityWith the Lakehouse Architecture on DatabricksVelu Nataranjan Principal Data Engineer,GoodRxYuga Chaparala Director,Data Engineering,GoodRxTHURSDAY MAY 7 3:00 3:45 P.M.DBTA DATA SUMMIT 2026 VELU NATARANJAN YUGA CHAPARALA#

2、DATASUMMIT1/30Agenda01GoodRx&DataOpportunitiesCompany context Current warehousearchitecture Scalabilityopportunities02ArchitectureEvaluationDesign principles evaluated architectures why open lakehousewins03What We BuiltMedallion Unity Catalog cross-engine interoperability04DataInteroperabilityOne ca

3、talog to unify governance that travelswith the data05Key TakeawaysWhat was harder whatsurprised us trade-offs open discussionDBTA DATA SUMMIT 2026 VELU NATARANJAN YUGA CHAPARALA#DATASUMMIT2/3001GoodRx&Data OpportunitiesCompany context Current warehouse architecture Scalability OpportunitiesDBTA DATA

4、 SUMMIT 2026 VELU NATARANJAN YUGA CHAPARALA#DATASUMMIT3/30W H O W E A R EGoodRx A Digital Healthcare PlatformMaking prescriptions more affordable and accessible for millions ofAmericans.Founded 2011.Serving customers in all 50 U.S.states.$17BSaved on prescriptions in 2024100M+Rx savings transactions

5、/yearWhat we doDiscounts&coupons price comparison across 70,000+pharmaciesTelehealth&subscriptions GoodRx Care+savings plansRx Smart Saver white-label savings program for health plansMedication delivery same-day&mail-order via ScriptDropMultiple lines of business One data platformDBTA DATA SUMMIT 20

6、26 VELU NATARANJAN YUGA CHAPARALA#DATASUMMIT4/30Where We StartedDBTA DATA SUMMIT 2026 VELU NATARANJAN YUGA CHAPARALA#DATASUMMIT5/30Scalability Opportunities$Cost at ScaleRedshift couples compute and storage payfor both even when only one is the bottleneck.*DS/ML InfraSQL warehouse,ML cluster two sys

word格式文档无特别注明外均可编辑修改,预览文件经过压缩,下载原文更清晰!
三个皮匠报告文库所有资源均是客户上传分享,仅供网友学习交流,未经上传用户书面授权,请勿作商用。
1. **GoodRx数据挑战**:原Redshift架构存在成本高(计算存储耦合)、实时性差(Kafka事件延迟)、治理分散(多副本)、厂商锁定(专有格式)问题。 2. **开放湖仓架构胜出**:基于8项设计原则(如存储计算解耦、ACID、统一治理),Apache Iceberg+Unity Catalog架构满分(8/8)优于传统仓库(4/8)和Glue湖仓(6/8)。 3. **核心成果**: - **零副本治理**:单表支持Databricks、Athena、Redshift Spectrum跨引擎查询,统一策略(掩码、审计)自动生效。 - **AI就绪**:受治理的Iceberg表直接用于ML训练,特征表无泄漏,模型注册表支持全链路追溯。 - **性能提升**:自动优化(ZORDER、液态聚类)使查询延迟降低20倍,审计时间从周级缩至单查询。 4. **权衡与进展**:Redshift集成依赖AWS预览功能(凭证分发),需增量迁移;ML特性(特征存储、向量搜索)需Delta层补充。
**湖仓架构优势?** **数据互通难题?** **治理如何落地?**
客服
商务合作
小程序
服务号
折叠