Overview
Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only!
Discover how to leverage RAPIDS Accelerator for Apache Spark and Alluxio to advance GPU analytics in this 27-minute talk from Databricks. Learn about RAPIDS, a set of open source libraries enabling GPU-aware scheduling and memory representation for analytics and AI, and how Apache Spark 3.0 utilizes it for GPU computing. Explore the need for accelerating data access with Alluxio to complement the massive parallelism of GPUs. Gain insights into using Alluxio and Spark with RAPIDS Accelerator on NVIDIA GPUs without application changes. Delve into trends in analytics and AI, modern data pipelines, data orchestration for GPU clusters, benchmarking results, and key configurations for Alluxio and RAPIDS Accelerator. Understand concepts such as data accessibility, filesystem namespace, metadata locality, and asynchronous data operations in the context of GPU-accelerated analytics.
Syllabus
Intro
TRENDS FOR ANALYTICS AND AIDATA PIPELINES
A MODERN DATA ANALYTICS OR AI PIPELINE
DATA ORCHESTRATION FOR GPU CLUSTERS WITHIN A SINGLE DATACENTER OR CLOUD
NVIDIA Brings GPU Acceleration to Apache Spark
Alluxio and RAPIDS Accelerator for Apache Spark
BENCHMARKING ENVIRONMENT
BENCHMARKING RESULTS
ALLUXIO CONFIGURATION
RAPIDS ACCELERATOR CONFIGURATION
DATA ACCESSIBILITY
ALLUXIO FILESYSTEM NAMESPACE
METADATA LOCALITY
ASYNCHRONOUS DATA OPERATIONS
References
Taught by
Databricks