From the course: NVIDIA Certified Associate AI Infrastructure and Operations (NCA-AIIO) Cert Prep

Unlock this course with a free trial

Join today to access over 25,300 courses taught by industry experts.

Network inside an AI-centric data center

Network inside an AI-centric data center - NVIDIA Tutorial

From the course: NVIDIA Certified Associate AI Infrastructure and Operations (NCA-AIIO) Cert Prep

Start my 1-month free trial Buy for my team

Network inside an AI-centric data center

“

So you now know about the compute aspect of AI-centric data center, that is CPU, GPU, and DPU. What about communication between these devices? That is where network is required. So when it comes to a network inside an AI-centric data center, there may be various types of network operating there. What do we mean by these various types of network? There could be network for compute, storage, in-band management, out-of-band management. So why we need separate network? Let's focus first on that, and then I'll explain you usage of each of these type of network. So we don't want to put all the traffic on one single network thread, because there would be a lot of issues because of that. So why the separation matters? Because we may want to have isolation for performance. So maybe my compute network and storage network would require a lot of bandwidth, But my inbound and outbound management may not require that much of bandwidth. So we may want to keep performance isolated as much as…

Contents