DCops AI
Login
Menu

Protected

Event details are available after login. Redirecting…

If you are not redirected, login.

Event

AII SME Webinar #04

Technical deep dive on NVIDIA BlueField DPU: where it fits in AI infrastructure, what to offload, and how to operate it in production clusters.

2026-03-07 · 9:00 AM PST · Remote

Why this webinar exists

As AI clusters scale, CPU cycles are increasingly consumed by networking, security, and storage-side operations.

BlueField DPUs shift critical data-path and infrastructure services away from host CPUs to improve throughput and determinism.

Many teams have hardware but lack clarity on where DPU offload delivers measurable operational value.

This session helps platform engineers map BlueField capabilities to real production bottlenecks and deployment choices.

Seats are limited to 200 per session. Each person can register for one session at a time. You can switch by unregistering.

What we will cover

  • BlueField DPU architecture in the AI infrastructure stack
  • Control plane vs data plane responsibilities
  • Networking, storage, and security offload patterns
  • Kubernetes and cluster integration strategy
  • Operational troubleshooting and observability model

Agenda

  1. DPU foundations: what BlueField is and where it fits
  2. AI cluster bottlenecks DPUs can offload today
  3. DOCA and software stack overview for operators
  4. Fabric path design: NVLink, PCIe, Ethernet/InfiniBand touchpoints
  5. Kubernetes integration patterns and day-2 operations
  6. Firmware, lifecycle management, and upgrade guardrails
  7. Failure modes and troubleshooting workflow
  8. Q&A: deployment strategy for enterprise and cloud clusters

Who should attend

  • Platform and infrastructure engineers
  • SRE / production engineering teams
  • Network and cluster operators
  • Architects designing GPU and data-path infrastructure

What this session is not

  • Introductory AI/ML theory
  • Application framework coding tutorials
  • Vendor marketing walkthroughs

This is a systems-level engineering session focused on practical DPU adoption in AI infrastructure.

Format

  • Duration: 60 minutes
  • Live Q&A: 15-20 minutes
  • Remote session
  • Seats limited to 200 per session

Duration: 60 minutes + Live Q&A

BlueField DPUs are becoming a core lever in high-performance AI infrastructure.

Teams that understand offload boundaries will operate clusters more efficiently.

Register to secure your seat.