Maximizing Programmable DPUs in the Network
summary
Describes a three‑year POC that integrates programmable DPUs into legacy telecoms to accelerate UPF-like functions, using DOCA API, C with CUDA plus Python and Go, containerized on Kubernetes. Two full‑time engineers build a DPU‑accelerated cUPF prototype targeting DGX for central and Bluefield‑3 for edge, run QoS tests that show strong performance, and evaluate multi‑tenancy and GPU‑as‑a‑Service with BIG‑IP Next for Kubernetes.
AI-written summary · about
related talks
39mFunction Dedicated Network Deployment Model in DPU/IPU-Like Device Acro...
30mZXDH DPU Adapter and It's Application
19mCluster Optimization: Harnessing DPUs for Ambient Mesh Workloads
43mWhat is the Big Deal with the Open Programmable Infrastructure Project?
18mSmartNIC DPU Enablement in OpenStack with OVN
13mLightning Talk: Accelerating Cilium: Our Journey on Offloading Cilium To DPU
18mArchitecting Multi-tenant Data-center Networks for GPU Customers
28mCeph RBD Integration with DPUs