Keynote: The Future and Value of Composable Disaggregated Infrastructure
summary
Composable Disaggregated Infrastructure disaggregates servers into pools of CPU, memory and accelerators connected by PCI or CXL 3.0 shared memory switches, enabling software‑defined custom servers and dynamic GPU attach/detach under Kubernetes. The demos show reduced inter‑service latency vs HTTP by using shared memory, adaptive power saving via dynamic device scaling, and integration work with Linux Kernel, Kubernetes and OPI. Vendor measurements claim up to 40% capex reduction and lower Opex from on‑demand resource composition.
AI-written summary · about
related talks
05mTowards a Sustainable World Realized by Kubernetes and Composable...
05mProject Lightning Talk: CoHDI: Dynamic Hardware Composition for AI‑Era Kubernetes…
41mDynamic GPU Provisioning in Kubernetes With Composable Disaggregated...
05mSponsored Keynote: The Cloud (And Your $$$) is Not Infinite: Dynamic Scheduling at Every Layer
15mKeynote: From Platforms to AI Factories. Has Kubernetes Solved It?
34mDeclarative Device Virtualization: Orchestrating GPUs & Hardware in Cloud Native Environments
24mUsing Kubernetes with Data Processing Units to Offload Infrastructure
08mContainerization – It’s Not Just for CPUs Anymore: Introducin...