This session is tailored for AI/ML and data practitioners aiming to build scalable AI/ML data pipelines and select the optimal storage solutions. Learn to optimize AI/ML workloads, including data preparation, training, tuning, inference, and serving, using the best storage options that seamlessly integrate with Compute Engine, Google Kubernetes Engine, or Vertex workflows. Explore how to enhance analytics workloads with Cloud Storage and Anywhere Cache.
Jason Wu, Director of Product Management for Google Cloud Storage, kicks off the session by outlining the challenges of defining a storage infrastructure for AI and analytical workloads. Joining him are David Stiver from Google Cloud, and Yuki Yachide and Alex Bain from Woven by Toyota, who share their experiences leveraging Cloud AI Storage.
The agenda includes an overview of AI workload challenges, the vision for cloud AI storage, and the latest product capabilities announced during Google Cloud Next. Woven by Toyota demonstrates their use of Cloud AI Storage on Google Cloud, followed by a Q&A session.
Jason highlights the critical role of storage in AI, noting that AI workloads are data-intensive and present unique challenges for storage infrastructure. He outlines a typical AI data pipeline, from data ingestion and preprocessing to model training, validation, and inferencing. The iterative nature of AI pipelines often requires better data for improved model accuracy, necessitating robust storage solutions.
The discussion delves into the four key requirements for AI storage infrastructure: performance and scalability, data management, data governance, and ecosystem support. Jason presents Google Cloud’s vision for AI storage: storing data once, accessible from anywhere, with any performance and interface, and at any scale. This vision simplifies data management, reduces costs, and enhances accessibility and performance.
David Stiver elaborates on five new products designed to accelerate AI/ML workloads, integrated into the AI data pipeline stages. These products include Anywhere Cache, Cloud Storage Fuse with local caching, Parallel Store, and Hyperdisk for ML, all aimed at improving data throughput, performance, and efficiency.
Yusuke Yachide and Alex Bain from Woven by Toyota share their practical experiences, highlighting the benefits of Cloud Storage Fuse in reducing training costs and improving performance. They discuss their challenges with previous storage solutions and how Google Cloud’s offerings have streamlined their processes.
To conclude, Jason encourages attendees to explore Cloud AI Storage and experience the performance and savings achieved by Toyota. He invites feedback and engagement through a QR code for connecting with the product team.
Speakers: David Stiver, Alex Bain, Jason Wu, Yusuke Yachide.
