FINRA：利用 NVIDIA 在 AWS EMR 上加速海量数据处理（由 NVIDIA 赞助）.pdf

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Felix CheungProduct Manager Accelerator for Apache SparkNVIDIAAlain MenguySenior DirectorBigData&P

2、erformance EngineeringFINRAFINRA:Accelerate Massive Data Processing with NVIDIA on AWS EMRA I M 2 7 9-S 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.What is NVIDIA RAPIDS Accelerator for Apache SparkWho is FINRA,and what is FINRAs mission?Life in the cloud:flywheel/Moores lawIn

3、creases in market volume can lead to rising compute costs and longer processing times of our SLA-driven Spark workloadsCan GPU acceleration via Spark RAPIDS can reduce costs and improve throughput?Key takeawaysHow will we spend our time today?2025,Amazon Web Services,Inc.or its affiliates.All rights

4、 reserved.Apache Spark is ubiquitous in Modern Enterprises80%of the Fortune 500 use Apache Spark 3.x1Object Store(Data Lake)Data WarehouseRelational DBAnalyticsMachine Learning PipelineData Analytics TeamBusiness Intelligence TeamAI/Data Science TeamRaw Data from Sales,Suppliers,Customers,Operations

5、Used broadly for ETL and Feature Engineering1 https:/spark.apache.org Source:IDC,Revelations in the Global DataSphere,US49346123,July 2023 Data Volume in Zetabytes0501001502002503003502017 2018 2019 20202021 2022 2023 2024 20252026 2027 2025,Amazon Web Services,Inc.or its affiliates.All rights reser

6、ved.NVIDIA RAPIDS Accelerator for Apache SparkImprove existing data processing workflowsAmazon EMRFaster ResultsTake advantage of high-performance analyticsAccelerate data modeling and ML/AI pipelinesLower CostsSave on infrastructureReduce power consumption,cooling and carbon footprintBenefitsCloudO