If a Duck Quacks In the ForestShould You Care？.pdf

1、Ryan BoydCo-founder MotherDuckryguyrgIf a Duck QuacksIn the Forest,Should You Care?!?#DuckTalk#DuckPost#DuckPointTHIS PRESENTATION IS NOTA complete tutorial on DuckDB THIS PRESENTATION DUCKING ISAn explanation of the industry context in which DuckDB came to lifeA discussion on OSS DuckDB and why its

2、 awesomeA spotlight on the internals of DuckDB and whyA demonstration of DuckDBs capabilitiesA preview of MotherDuck and its architectureLicensed from AdobeIntroduced Computing Licensed from AdobeThe Good Powerful Centralized Compute Shared Access The Bad Bulky Expensive Not Personalized Licensed fr

3、om AdobePopularized Computing Dedicated Personalized Why?DALL-E generated imageCDWDALL-E generated image32 GB RAM10 CoresDALL-E generated imagecreated at:created by:maintained by:Community&FoundationLets Break that DownIn-ProcessRuns inside Python,R,C+.No external dependencies.ApplicationDuckDBParse

4、BindOptimizeExecuteresultsSQLLets Break that DownIn-ProcessRuns inside Python,R,C+.No external dependencies.OLAPFull SQL support-things like window functionsHigh-performance aggregationsSQLite for Analyticscreated at:created by:maintained by:Community&FoundationWHY?!*High Performance Architecture*Si

5、mplified Data Access*OSS Community Flocks Together*Fast-paced innovationHigh Performance ArchitectureComparing Database EnginesRow-based(tuples)ColumnarVector-basedSQLite,PostgreSQL,etcPandas,NumPy,etc.DuckDBComparing Database EnginesRow-based(tuples)SQLite,PostgreSQL,etcOptimized for:*low memory fo

6、otprint*transactional workloadsComparing Database EnginesColumnarPandas,NumPy,etc.Optimized for:*analytic workloads*aggregations*data compression on like dataComparing Database EnginesVector-basedDuckDBOptimized for:*analytic workloads*aggregations*CPU-can do SIMD*CPU-cache localityWhy Vectorized?CP