Join us
Updates and recent posts about Slurm..
Gateway API Inference Extensiontakes AI workload routing on Kubernetes and infuses it with model-savvy powers. It slices latency on GPU clusters like a samurai. Meanwhile, theEndpoint Selection Extensionacts like a traffic cop on caffeine, using live metrics to steer pods and trim those nagging tail.. read more
Karpenter 1.0is the speedy barista of Kubernetes. It whips up nodes on demand, slashing AWS EC2 costs by 30-50%. Why? Real-time scaling magic, Spot instance wizardry, and APIs that won't stab you in the back. Sure,Cluster Autoscalerhas an extensive resume of compatibility and control, but it's like .. read more
Omio swapped Spot VMs for standard ones in a single region and unearthed a shocker. Costs didn't skyrocket; they actually dropped. Network glitches? Gone. They braced for a70% budget implosionbut emerged with a grin. Standardizing on 16-core, 1:4 RAM machines cranked up performance and dialed down c.. read more
This post walks through an example of combining a FastAPI service with an eBPF tracer to monitor syscalls. It covers common pitfalls encountered during development on macOS, the shift to containerizing the environment, and how the author ultimately succeeded in capturing the desired syscalls—a hands.. read more
Amazon EKS Dashboardtames the Kubernetes chaos with finesse. It brings all your clusters into one sharp, centralized view on AWS. Sprawl, security snags, ballooning support costs—gone in a flash. Assess upgrade needs, peek into cost forecasts, and manage add-ons without breaking a sweat. Wave farewe.. read more
Most projects don’t need Kubernetes;for 90% of teams, it adds unnecessary complexity and operational burden compared to simpler alternatives like managed cloud services, VMs, or actor-model frameworks. Unless you’re running at hyperscale, need true hybrid environments, or have a massive, mature plat.. read more
Tetragontaps into the kernel usingeBPF, giving containers an all-access pass without the agent baggage. When you pair Tetragon with AKS, you unlock crystal-clear views of process executions and system calls. Security teams revel in this treasure trove, primed for spotting and squashing threats swift.. read more
Meet the motley crew that is theYOLOv8-based AI team. These guys get serious about detecting hard hats across countless video streams and they do it in real time. Their secret weapon? The metallic trio ofZEDEDA,Rancher, andTerraform.ZEDEDAtames edge management.Rancherwrangles Kubernetes.Terraform? I.. read more
Terraformisn't just making waves atSCHIP; it's rewriting the rulebook. Watching CI plan times dive from a sluggish 10 minutes to a snappy 30 seconds feels like magic, thanks to its knack for spitting out import statements like they're hotcakes. While flashy automation dazzles, it's actually the grit.. read more
Just launched agent-manifest.org — a schema for describing what AI agents do, what they need, and how they work.
Agents are the new APIs
Agents are becoming the next layer of software abstraction—autonomous tools that act on our behalf, perform tasks, make decisions, and interact with APIs, data, and humans.
But as agents proliferate, we face a growing challenge:
How do we understand what an agent does, what it needs, and what it can be trusted with?
See the proposed standard to solve this problem
