Containerization – It’s Not Just for CPUs Anymore: Introducin...
Talkat Platform Engineering Day NA 2024 · on Tuesday, 2024-11-12· at 10:15· Salt Palace | Level 1 | Grand Ballroom G·
summary
Vultr claims a unified CPU+GPU Kubernetes platform running in 32 data centers with sub‑40ms reach to 90% of the world and a 100% uptime guarantee, using Cluster API and Crossplane to provision infrastructure, deploying workloads with Helm with NVIDIA/AMD operators, and exposing GPU fabric via RDMA so Kubernetes pods can run distributed training. Demo shows manifests that scale from one GPU to eight GPUs across regions, and registries hold immutable models and containers for global rollout.
AI-written summary · about
related talks
06mAutoscaling GPU Clusters Anywhere — Hyperscalers, Neoclouds & Ba...
30mExplain How Kubernetes Works With GPU Like I’m 5
30mRunning AI Workloads in Containers and Kubernetes
39m✨ GPU in My Containers!
30mBeyond the DC Walls: Building a Nationwide GPU Fabric With KubeVirt and Wide-Area…
12mBuilding Bulletproof AI Platforms
47mCloud Native and Sustainable, Reproducible Scientific Computing
34mDeclarative Device Virtualization: Orchestrating GPUs & Hardware in Cloud Native Environments