Class Central is learner-supported. When you buy through links on our site, we may earn an affiliate commission.

YouTube

Evolving Storage Solutions for AI/ML Performance and Scale

SNIAVideo via YouTube

Overview

Explore how storage systems are evolving to meet the demands of modern AI/ML workloads in this 36-minute conference talk from the Storage Developer Conference 2022. Discover Samsung's innovative approach to addressing storage challenges through specialized object storage solutions designed for machine learning applications. Learn about high-performance S3 capabilities, the open-source Samsung DSS stack's ability to handle intensive AI training storage requirements, and methods for identifying performance bottlenecks across different storage backends. Gain insights into DSS Enhanced Minio Object-Store implementation, reference deployment models, and comparative performance analysis between BeeGFS and DSS Gen2 systems. Through detailed benchmarking demonstrations and real-world examples, understand how combining optimized software and hardware architectures enables unprecedented storage performance and scalability for AI/ML workflows.

Syllabus

Intro
DSS: Performant & Scalable Object Storage
DSS Enhanced Minio Object-Store
Reference Minio + DSS deployment model for AMD
DSS Client Wrapper
DSS Deployment View
DSS GET Performance
Al Benchmarking Tool
Benchmarking on the pipeline
Reproduced the Baseline numbers in MSL Lab
Benchmark Topology and Configuration (DSS)
BeeGFS vs DSS Gen2 - DGX Client Peak BW Tests

Taught by

SNIAVideo

Reviews

Start your review of Evolving Storage Solutions for AI/ML Performance and Scale

Never Stop Learning.

Get personalized course recommendations, track subjects and courses with reminders, and more.

Someone learning on their laptop while sitting on the floor.