Leven Lake

Google met à jour Veo : la génération vidéo gagne en contrôle et en réalisme. moncarnet.com/2025/10/16/g...

Kubernetes 1.34: Security, Performance, and DRA Go GA - The Landscape Vyom Yadav, Kubernetes Release Team Lead and Software Engineer at Canonical, joins Sylvain Kalache to discuss what’s new in Kubernetes 1.34. With over 58 enhancements, this release focuses on maturing...

What does Kubernetes 1.34 say about the future of infra?

In The Landscape, Vyom Yadav breaks down upgrades in security, performance, and resource management.

CEL, admission policies, pod identity—it’s all in the details.

thelandsca.pe/2025/10/15/k...

#Kubernetes #CNCF #CloudNative #TheLandscape

OVHcloud met l’IA au service d’un refroidissement écoresponsable dans ses centres de données pour réduire la consommation d’eau de 30 % et celle d’électricité de 50 %. moncarnet.com/2025/10/16/o...

Ep 140 Shorts: Introduction to llm-d Open-source K8s-native Framework for Distributed LLM Inference YouTube video by Cloud Native Podcast

LLMs are monoliths, which can be a major cause for your CPU/GPU compute bills 📈. What if we can build a K8s-native distributed inference stack that brings cache-aware routing and disaggregated serving to LLMs?

Weclome LLM-D which does that, make ur compute bills 📉.
www.youtube.com/shorts/rI8zF...

Introduction to llm-d Open-source Kubernetes-native Framework for Distributed LLM Inference | Ep 140 YouTube video by Cloud Native Podcast

llm-d is a new opensource tool and approach designed to make serving generative models on K8s efficient, scalable, and cost-effective by introducing cache-aware routing, disaggregated serving (pre-fill/decode), and K8s-native scheduling & gateways.

🎧 to #CloudNativeFM 👇 youtu.be/2Wtug1kTwUk

Infinite Scroll

Demystifying Automatic Instrumentation: How the Magic Actually Works Despite the rise of OpenTelemetry and eBPF, most developers don’t know what automatic instrumentation actually does under the hood. This post breaks it down—not to suggest you build your own, but to…

Automatic instrumentation can seem like magic—but it’s not!

The latest #OpenTelemetry blog breaks down how it really works, from monkey patching and bytecode instrumentation to eBPF and runtime APIs.

buff.ly/aWbGOzf

Autoscaling vLLM with OpenShift AI | Red Hat Developer Implement cost-effective LLM serving on OpenShift AI with this step-by-step guide to configuring KServe's Serverless mode for vLLM autoscaling

Autoscaling vLLM with OpenShift AI | Red Hat Developer developers.redhat.com/articles/202...

Opération Rubicon : espionnage à l'échelle mondiale | RTS YouTube video by RTS - Radio Télévision Suisse

J'ai visionné il y a quelques jours un reportage de la RTS qui traite d'une affaire d'espionnage à l'échelle mondiale qui a concerné une société suisse, leader mondial du chiffrement "Crypto".

Je conseille à tout le monde : youtu.be/Wm1Vk90tUKw?...

What’s new in Flux 2.7 (including the External Artifacts API, Source Composition, Source Watcher), demo gitless GitOps with OCI artifacts, show performance & monitoring tooling, and use the Headlamp plugin to watch reconciliation in real time.

Coming Soon!!!! www.youtube.com/shorts/R86lh...

Kubernetes v1.34: Pods Report DRA Resource Health-

Kubernetes v1.34: DRA Consumable Capacity-

Powered by JetLag®

Kubernetes v1.34: DRA has graduated to GA-

Optimize GPU utilization with Kueue and KEDA | Red Hat Developer As GPU demand grows, idle time gets expensive. Learn how to efficiently manage AI workloads on OpenShift AI with Kueue and the custom metrics autoscaler

Optimize GPU utilization with Kueue and KEDA | Red Hat Developer developers.redhat.com/articles/202...

Cilium 1.18 - Expanded IPv6 Support, Encrypted Overlay, Ingress Bandwidth Controls, Policy Performance Improvements, and More!

Cilium 1.18 release blog is out now. My top two are support for IPv6 kube-proxy replacement and the performance improvements (reduced policy latency 40%, CPU usage down 43% under service churn, and 30% smaller arm64 images)

Let’s shape the future of LLM infrastructure on Kubernetes, together.

👉 Join a SIG. Bring your expertise. Build something that lasts. https://llm-d.ai/docs/community/sigs

Here we are - what can we do for you ?

🚀 Introducing @kubefloworg.bsky.social Trainer 2.0 — the next evolution in AI model training on @kubernetes.io!

We’re excited to announce 𝗞𝘂𝗯𝗲𝗳𝗹𝗼𝘄 𝗧𝗿𝗮𝗶𝗻𝗲𝗿 2.0 — tailored to simplify and scale AI model training in K8s-native environments.

🔍 What’s New in 2.0?

Les produits Microsoft sont devenus un aspirateur centralisé des données d'entreprise. Ne pas réagir est une faute stratégique historique pour les DSI.

Learn how to build a flexible GenAI platform using the open-source solutions Envoy AI Gateway, KServe, & complementary tools
- Self-Hosted Model Serving w/KServe
- Observability, Control, and Optimization for Prod Readiness
- Policy Enforcement and Guardrails
aigateway.envoyproxy.io/blog/envoy-a...

Pour avoir regardé quelques émissions de télé. L'équipe explose plutôt rapidement.

Maybe canadians impact on the top three.

Je préfère ce nouveau nom. Excellente nouvelle

Introducing Gateway API Inference Extension Modern generative AI and large language model (LLM) services create unique traffic-routing challenges on Kubernetes. Unlike typical short-lived, stateless web requests, LLM inference sessions are often...

Introducing Gateway API Inference Extension-

Son principe de pédagogie était que s’il ne pouvait présenter un sujet durant un cours de première année, c'est que lui ne l'avait pas complètement compris. Feynman eut beaucoup de plaisir à présenter son explication de « niveau première année » de la connexion spin-statistique quantique

Le plus bel exemple Richard Feynman. « Great explainer » (« le grand explicateur ») en effet, il prenait beaucoup de soin dans ses explications aux étudiants, en mettant un point d'honneur à ne pas utiliser de formulations pédantes, mais à être le plus accessible possible aux autres.

Kubernetes doesn’t make egress easy, scattering it across NAT tables, host routing rules, and CNI quirks. Teams end up reaching for hacky solutions like standalone proxies, policy engines, even hand-configured nodes

Enter stand Alone Egress Gateway

isovalent.com/blog/post/is...

We are excited to announce KServe v0.15 release, marking a significant leap forward in serving both predictive and generative AI models.

GenAI features: Envoy AI Gateway integration, multi-node inference via vLLM, LLM autoscaler, distributed KV cache via LMCache.

kserve.github.io/website/mast...

ML Engineers often struggle with inconsistent packaging mechanisms, forcing them to repackage models multiple times, slowing down development and increasing risk.

@Kit_Ops is solving these issues by standardizing model packaging & deployment.

🎧 to know more #cloudnativefm -> youtu.be/BM9PcoK2Ik8