释放 Iceberg 的力量：我们在 Databricks 上打造统一 Lakehouse 的旅程.pdf

1、Unlocking the Power of IcebergOur Journey to a Unified Lakehouse on DatabricksTomer Sabag10/6/2025Timeline32012Beginning2015Moving to Cloud2021New CEO2022HypergrowthLets goData-driven!2024DataFlexibility2025Speed&ScaleAbout MeLeading Platform Engineering at LSports10+years building real-time data pl

2、atforms in small/medium sized start-upsPassionate about explainable,trustworthy dataFocused on bridging product vision and technical execution4Tomer Sabag Director of Platform Engineering at LSportsLSports The BeginningBootstrap Ashkelon based company4 Customers2 RobotsBasic features:Automated Settl

3、ementsOdds Feeds5Real-time Sports Betting Data ProviderIdo LazarDotan LazarShaul LazarOur Cloud Journey BeginsNew IT and DevOps teamsEU based cloud,closer to our customers endpointsFaster development,faster deployments6Kicking servers out of Ashkelon hello AWSTimeline72012Beginning2015Moving to Clou

4、dBuilding Our Data CultureDotan steps in as CEONew CDO and BI departmentMandate:build centralized data visibility and decision-making power AWS Redshift stackEnabled dashboards and KPIs for the first time8No more gut feelings lets get data-drivenOur Redshift Architecture9Timeline102012Beginning2015M

5、oving to Cloud2021New CEO2022RedshiftChallenges We Faced with RedshiftLong vacuum processesDashboard timeoutsCoupled storage&computeLimited size of saved dataOptimizing expensive SQL queriesPerformanceCostsManaging partitionsHandling schema evolutionMaintaining our own batch processing componentDev

6、Overhead11Slow queries,rising costs,and a ceiling we couldnt breakSearching for Our Next SolutionOpen format growing community,multi-engine supportFeatures schema evolution,time travel,hidden partitionsDecoupling compute and storage Partnered with Tabular for managed Iceberg catalog12Our love for op