2115 - Meet the Future of Storage for Data and AI.pdf

《2115 - Meet the Future of Storage for Data and AI.pdf》由会员分享，可在线阅读，更多相关《2115 - Meet the Future of Storage for Data and AI.pdf（27页珍藏版）》请在三个皮匠报告上搜索。

1、October 21-24,2024Mandalay Bay Convention CenterLas Vegas,NevadaSession code 2115Danny Mace,VP Storage Software Development,IBM StorageAlbert Ho,VP Strategy and Product Management,IBM StorageMeet the Future of Storage for Data and AI Keynote,Directions and StrategyAgenda01020304IBM Storage for Data

2、and AIModernizing Data Lakes with Storage CephOptimize training with IBM ScaleTraining and Inferencing with IBM FusionIBM TechXchange|2024 IBM CorporationTheres more dataExploding data growthThe aggregate volume of data stored is set to grow over 250%in the next 5 years.In more locationsMultiple loc

3、ations,clouds,applications and silos82%of enterprises are inhibited by data silos.In more formats Documents,images,video80%of time is spent on data cleaning,integration and preparation.With less quality Stale and inconsistent82%of enterprises say data quality is a barrier on their data integration p

4、rojects.3Storage teams are faced with unprecedented data data challenges challenges to scale AI Leading to more cost and complexities with governing data used for AIFeb 2024 PLT-Storage Software Storage infrastructure plays a foundational role in your AI strategyStorage infrastructure plays a founda

5、tional role in your AI strategyDataDecisionsInferenceModeladaptationDistributed training&model validationMay have sensitivity to latency/throughput,always cost-sensitive Long-running job on massive infrastructureDatapreparationWorkflow of steps(e.g.deduplicate,remove hate&profanity,etc.Model tuning

6、with custom data set for downstream tasksPreparePrepareBuildBuildDeployDeployHours to daysWeeks to monthsMinutes to hoursSub-second API requestLanding Zone for High-Speed Data Ingest High Performance Storage for Preparing AI Models:Data LakehouseHigh Performance Storage for Training AI Models:High P