This article is more than 1 year old
Want to beef up your GPU ops? You’ll need to bulk up your storage too
Play catch up then get ahead
Webinar If GPUs have revolutionized AI and HPC over the last decade, it's only because they could rely on storage infrastructure capable of delivering massive amounts of data.
Which means that if you're staking your future on GPUs, you need to have your storage strategy nailed down as well.
So, what is the state of storage today? What effect will the hybrid cloud era have? And where do newer technologies like DPUs and NVM Express over Fabric (NVMe-oF) fit in?
You can get on top of all these questions by catching up on this Open Storage Summit Session, Feeding the Data-Hungry GPU Beast.
This session sees the Register's Tim Prickett Morgan joined by Rob Davis, Vice President of Storage Technology at NVIDIA, and Randy Kreiser, Storage Specialist/FAE at Supermicro.
They talk through the role of the programmable, SoC-based DPU in the evolution of datacenter HPC and AI systems, plus the development of NVMe-oF and how it is expected to change the way storage is architected for HPC and AI systems.
They also answer the burning practical questions of the day, such as how different organizations' priorities around performance and data types affect their file system selection. The trio go on to demonstrate how GPUDirect Storage can boost storage performance as well as AI and HPC apps, and how you can optimize connectivity between the storage nodes and clusters.
If that's whetted your appetite, you'll be glad to know this is just one of many sessions making up Open Storage Summit, brought to you by Supermicro, which bring together data specialists, and datacenter application professionals. And they're all available, right now.
Whether you're looking to understand the state of the art when it comes to high performance storage, find practical tips on optimizing your own infrastructure, or work out where the industry is going in the next few years, head here and peruse the agenda.
Sponsored by Supermicro.