Strategic PartnersIssue 665 records
NVIDIA
A KubeCon-driven edition on Kubernetes-based AI factory software and independent scrutiny, with no new SLED deployment evidence. NVIDIA contributed OpenShell to the CNCF Sandbox and announced NVKE and NVKR (public preview, not recommended for production) and Snapshot, which it reports restores GPU inference replicas 8x to 36x faster than cold starts on B200 with VAST storage. Independent reporting stresses that Sandbox is an entry tier and not a security audit, and an Oct 30 independent benchmarking paper finds B300 is not a uniform successor to B200, with a development vLLM build alone changing time-to-first-token by 2.1x. The common thread is that preview-stage tooling and vendor-measured results need local, workload-specific validation.
- Evidence records
- 5
- Cross-source patterns
- 2
- Evidence classes
- 3 vendor claim1 independent reporting1 academic research
- Outcomes
- 3 emerging1 cautionary1 mixed
- Source freshness
- 5 new this fortnight
- Research completed
- 2026-11-11
Choose a role to see its takeaway beside every record in the ledger.
Patterns across the evidence
Synthesis · Lighthouse Advisory interpretation
New open source AI-factory and agent-security tools arrive at preview or sandbox maturity
NVIDIA's KubeCon announcements (OpenShell to CNCF Sandbox, NVKE and NVKR in public preview) come with partner integration plans rather than shipping products. The vendor blog says NVKE and NVKR are not recommended for production, and independent reporting notes Sandbox is an entry tier, not a security audit, and that no independent review of OpenShell isolation is stated.
Operating questionWhat maturity, support and independent assurance evidence will we require before moving an open source preview or sandbox project into production?
Supporting evidenceNVIDIA Expands Open Source Commitment With OpenShell Donation and New AI InfrastructureManage Kubernetes Clusters for GPU Infrastructure with NVIDIA Kubernetes EngineAIholics
Measured GPU results depend heavily on software build, storage and workload
NVIDIA's Snapshot post reports 8x to 36x faster restores on B200 with VAST NFS and specified engine versions, and notes dependencies such as same GPU and driver and exclusion of image pulls. An independent benchmarking paper finds B300 is not a uniform successor to B200 and that a development vLLM build alone changed time-to-first-token by 2.1x between identical-silicon nodes.
Operating questionWhich of our own workloads, software builds and storage configurations must we benchmark before relying on published GPU performance figures?
Supporting evidenceStart and Restore GPU Workloads in Seconds with NVIDIA SnapshotMohsen Zand, Trung Dac Nguyen, H. Birali Runesha
Evidence ledger
Full record · every source keeps its link and limitations
NVIDIA Expands Open Source Commitment With OpenShell Donation and New AI Infrastructure
At KubeCon, NVIDIA said it is contributing OpenShell, its open source runtime for securing and governing autonomous agents, to the CNCF as a Sandbox project, and contributing to Agent Substrate, a Google-led CNCF Sandbox framework for long-running agent workloads on Kubernetes. It also announced three DSX AI Factory Platform components: NVKE and NVKR (public preview) and Snapshot. NVIDIA says hundreds of organizations have used OpenShell and that it has received tens of thousands of contributions in under a year. An OpenShell and Agent Substrate integration with Google is described as a collaboration, with no date given.
Why it matters, evidence and limitations
- Why it matters
- Indirect but real for Government and Education teams evaluating AI agents: policy enforcement outside the agent is a control point that security and procurement reviewers ask about. The post gives no SLED deployment evidence, so the relevance is architectural rather than proven.
- Evidence and measured results
- Vendor blog read in full. Adoption figures (hundreds of organizations, tens of thousands of contributions) are NVIDIA statements with no independent verification in the post. Partner and integration names for NVKE/NVKR are described as aligned on integration plans, not shipped products.
- Limitations and uncertainty
- Vendor announcement. Sandbox is CNCF's entry tier. No benchmarks, security audit or SLED deployment evidence in this post. Integration with Agent Substrate has no stated date.
Manage Kubernetes Clusters for GPU Infrastructure with NVIDIA Kubernetes Engine
NVIDIA announced two open source projects, NVIDIA Kubernetes Engine (NVKE) and NVIDIA Kubernetes Recipes (NVKR), for running managed Kubernetes services on GPU infrastructure. NVKR builds version-pinned node and container images and records build inputs; NVKE provisions and manages clusters through a REST API and controllers, with servers supplied by NVIDIA Infra Controller. Both are Apache 2.0, in public preview, and NVIDIA states they are not yet recommended for production workloads. NVKE is described as a Certified Kubernetes AI Platform and Installer, conformance tested on Kubernetes 1.36. Firmus and SK Telecom are integrating them; Mirantis, Rafay, Spectro Cloud, OpenNebula Systems, vCluster and Netris are incorporating components.
Why it matters, evidence and limitations
- Why it matters
- Limited and indirect. This targets AI factory operators and cloud providers, not typical agencies or campuses. It matters to SLED teams only where a state, university or research-computing center operates its own GPU cluster or buys capacity from a provider that adopts it.
- Evidence and measured results
- Vendor technical post with a walkthrough; no independent test, customer deployment or performance data. Partner integrations are stated as in progress.
- Limitations and uncertainty
- Vendor-authored; public preview; not recommended for production; walkthrough prerequisites are substantial (multiple CPU servers, GPU pool, NICo installation, build host).
Start and Restore GPU Workloads in Seconds with NVIDIA Snapshot
NVIDIA Snapshot is an open source checkpoint and restore system for GPU workloads on Kubernetes. On NVIDIA B200 GPUs with VAST NFS storage, NVIDIA reports restores of 3.6 to 42.7 seconds, 8x to 36x faster than cold starts of about 87 seconds to 18.5 minutes across vLLM, SGLang and TensorRT LLM, including models with checkpoints up to about 1 TB over eight GPUs. Checkpoint creation took 15 seconds to 5.5 minutes, and checkpoint plus restore was shorter than one cold start in every tested configuration. Design partners report Crusoe cutting GLM-5.3 readiness from about eight minutes to 31 seconds and Saturn Cloud cutting vLLM startup from 2 to 8.5 minutes to 6 to 34 seconds, both on HGX B300.
Why it matters, evidence and limitations
- Why it matters
- Limited and indirect. Faster inference scale-up and recovery matters to operators of shared GPU services, such as university research computing or state-run AI platforms, but the post offers no SLED deployment evidence.
- Evidence and measured results
- Vendor post with detailed tables for one GPU type (B200), one storage backend (VAST NFS) and specified engine versions; excludes container image pull and provisioning from the measured path. Partner numbers are summarized without test conditions in the post.
- Limitations and uncertainty
- Vendor-run benchmarks on one GPU model and one storage backend; same-GPU-and-driver restore requirement; workloads must meet compatibility requirements; image pull and provisioning time excluded; not independently reproduced.
Nvidia hands its AI agent sandbox OpenShell to the CNCF
Independent reporting on the same-day OpenShell donation says the move changes who steers the project, not whether developers can use it today, and that CNCF Sandbox is an entry tier for experimental projects, signaling neutral hosting rather than a security audit. It notes NVIDIA's adoption figures are its own and unverified, that NVIDIA does not say whether any independent review of OpenShell's isolation has occurred, that no date or shipping form is given for the OpenShell and Agent Substrate integration, and that NVKE/NVKR partner integrations are plans rather than shipping products. Partner Snapshot numbers are described as single-hardware-generation benchmarks without stated model sizes or test conditions.
Why it matters, evidence and limitations
- Why it matters
- Useful context for Government and Education teams weighing OpenShell: it separates governance neutrality from security assurance, which matters in agency risk reviews.
- Evidence and measured results
- Trade-press article read in full; it relies largely on NVIDIA's announcement and NVIDIA's OpenShell documentation, with no original testing reported.
- Limitations and uncertainty
- Secondary reporting derived mostly from vendor material; no hands-on testing; author-level independence from NVIDIA is not established beyond editorial framing.
Matching Workloads to Blackwell: Power-Resolved Benchmarking of NVIDIA B200 and B300 across HPC and AI Workloads
On two independent eight-GPU clusters, the authors find the B300 (Blackwell Ultra) is not a uniform improvement over the B200. It trails by up to 5.6x on convolutional inference, 8x on 3D vision transformer training, 24 to 27% on 70B LLM throughput and 16x on double-precision Quantum ESPRESSO, while winning on ViT-Large. On affected workloads the B300 draws only 36 to 53% of its 1100 W envelope, which the authors use to localize the cause to per-kernel cost rather than data supply. Two B300 nodes of identical silicon differ by 2.1x in time-to-first-token solely from a development vLLM build. They recommend per-workload procurement decisions benchmarked on the exact software build.
Why it matters, evidence and limitations
- Why it matters
- Directly useful to university research computing and public-sector HPC buyers choosing between Blackwell parts: it supplies independent, workload-level evidence and a method to test before buying. It reports no SLED deployment outcome.
- Evidence and measured results
- Abstract page read, plus the full HTML text for method and limitations. Harness, raw CSVs and plotting scripts are released. The paper states replication is uneven: CV and LLM results are single runs with spot-checked 2 to 4% variation; 3D ViT and HPC results are medians of five runs.
- Limitations and uncertainty
- Preprint accepted at a workshop; single-run CV and LLM measurements; harness keeps inputs in device memory so input-pipeline effects are excluded; cross-generation comparisons mix compiler and MPI versions; hardware access was provided by NVIDIA, Schmidt Sciences and Lenovo per the acknowledgments; the authors disclose LLM assistance in prose. Published October 30, newly relevant on Nov 10 because NVIDIA's same-day Snapshot post reports B200 and B300 results, and the paper shows how build and platform choices change measured outcomes.
How to read this edition
Source findings, measured results and limitations come from the cited publications. Patterns, operating questions, role takeaways and implementation considerations are Lighthouse Advisory interpretation, stated as questions to validate locally rather than guaranteed outcomes. Vendor and operator claims are labeled as claims. Full research method.
- Vendor claim
- A supplier-provided assertion that has not been upgraded to independent evidence.
- Independent reporting
- Independent reporting with attributable sources but without a formal evaluation design.
- Academic research
- Research produced through an academic institution or peer-reviewed venue.