Historically, Kubernetes managed accelerators as basic numbers, forcing workloads to claim entire physical cards regardless of actual compute requirements. Red Hat OpenShift 5 solves this by using dynamic resource allocation to inspect detailed device profiles and match incoming jobs via explicit ResourceClaims. Under this architecture, an NVIDIA H100 80 GB GPU can expose both whole-device access and fractional slices without triggering hardware conflicts. Meanwhile, a dedicated DeviceClass maintains backward compatibility by automatically converting legacy integer requests into modern claims without requiring manifest rewrites.
Key points
Workload scheduling now relies on precise hardware profiles described directly within dyn…: Workload scheduling now relies on precise hardware profiles described directly within dynamic ResourceClaims.
Shared counter sets prevent the scheduler from creating conflicting allocations between f…: Shared counter sets prevent the scheduler from creating conflicting allocations between full GPUs and smaller sliced instances.
Automatic conversion by the DeviceClass ensures existing application manifests function p…: Automatic conversion by the DeviceClass ensures existing application manifests function properly without needing manual rewrites.
Sources checked by our analysis pipeline: 1
For years, asking Kubernetes for a GPU has meant writing 1 line: nvidia.com/gpu: 1
. The value requested in that line is an integer. The scheduler knows how many GPUs a node has and how many a pod wants, and nothing else. It can't tell which GPU you get, how much of its memory you need, whether a slice of it would do, or whether the card had a fault 10 minutes ago. You get the whole card or nothing.
Dynamic resource allocation (DRA) changes the model. A DRA driver publishes every device as an API object with attributes and capacity. Workloads describe what they need in a ResourceClaim. The scheduler, not a node-local plugin, matches them. Red Hat OpenShift 5, built on Kubernetes 1.36, ships 6 DRA features on top of that model. The clearest way to show what they do is to watch them work together on a single card.
Red Hat OpenShift lets teams manage AI workloads through familiar Kubernetes APIs and operational workflows. DRA adds richer ways to request and allocate accelerators, helping platform teams match…
Routine automated checks inadvertently knocked servers offline, triggering an infrastructure overload that severed critical Microsoft 365 services across more than 70 countries. The widespread disruption crippled government agencies, judicial operations, and commercial communications for nearly an entire day. This systemic failure has intensified scrutiny from experts regarding the vulnerabilities of concentrating critical global infrastructure within a few dominant technology firms.
Weighing only 1.5 ounces, the new CS-8 Micro introduces an ultra-portable design to Camp Snap's vintage-inspired line while trading high video fidelity for pocket convenience. Unlike previous screenless devices from the brand, it adds a dedicated electronic viewfinder, though users must still transfer files to an external computer or smartphone to view their clips. This launch aims to capture the nostalgia market through a blind-box distribution model with low-friction retro video recording.
The open-source office suite project released a direct public statement emphasizing that its software operates without interruption year-round. This public communication underscores the platform's continuous operational availability for its global user base.