吴英骏-打造云原生流数据库（GOTC深圳会场）（27页）.pdf

1、分布式数据库与存储专场吴英骏 2021年08月01日本期议题：打造云原生流式数据库About MeYingjun Wu(吴英骏)CEO of Singularity Data Inc.Former software engineer at AWS RedshiftFormer researcher at IBM Research-AlmadenPh.D.,National University of SingaporeVisiting Ph.D.,Carnegie Mellon UniversityCo-founderHuanchen ZhangAssistant professor at T

2、singhua Yao classFormer postdoc fellow at Snowflake Inc.Ph.D.,Carnegie Mellon UniversityWinner of Jim Gray Dissertation Award(2021)About Our CompanySingularity Data Inc.10+full-time engineers with solid background from big techsRaised millions of U.S.dollars from top-tier VC firmsBased in Beijing,Sh

3、anghai,and the San Francisco Bay AreaExisting Cloud Data WarehousesStarting from Redshift/Snowflake.5-10 times larger than Snowflake Biggest software IPO in the history(once reached 80B+market cap)Existing Cloud Data WarehousesStarting from Redshift/Snowflake.Data scienceReportingLog analysisExplora

4、tory applicationsCloud data warehousesExisting Cloud Data WarehousesBut what if supporting streaming applications?Streaming applicationsGamingMonitoringAds bidding/recommendation?Cloud data warehousesExisting Cloud Data WarehousesBut what if supporting streaming applications?Streaming applicationsGa

5、mingMonitoringAds bidding/recommendationPoor performanceCloud data warehousesExisting Cloud Data WarehousesBut what if supporting streaming applications?Understand the workload characteristicsRevisit Redshift/Snowflakes architectureWorkload CharacteristicsWhy streaming applications are different?Hig

6、h ingestion rateLow query latencyContinuous computationCloud Data Warehouse ArchitectureCloud Data Warehouse ArchitectureHigh S3 commit latency results in low ingestion rate Cloud Data Warehouse ArchitectureHigh S3 commit latency results in low ingestion rate Shared computation resource results in p