Blog
1938+ articles — Page 65 of 81
Terraform vs Ansible
Terraform for provisioning, Ansible for configuration? Compare use cases, strengths, and when to use both together in your infrastructure pipeline.
2 min readTerraform vs ARM Templates
Moving from ARM Templates to Terraform? Compare readability, modularity, state management, and migration strategies. Practical guide with examples, commands.
2 min readTerraform vs Bicep
Terraform vs Azure Bicep for Azure infrastructure. Compare syntax, Azure-native features, state management, and multi-cloud capabilities.
2 min readTerraform vs CDK
Terraform HCL vs AWS CDK with TypeScript/Python. Compare developer experience, abstraction levels, and deployment workflows.
2 min readTerraform vs OpenTofu
Terraform vs its open-source fork OpenTofu. Compare licensing, features, community support, and migration paths between the two tools.
2 min readTerraform vs Pulumi
Which IaC tool is right for you? Compare Terraform's HCL declarative approach with Pulumi's general-purpose language support, state management, and ecosystem.
2 min readTerraform vs Terragrunt
Terraform alone vs Terraform with Terragrunt. Compare DRY configurations, multi-environment management, and dependency handling.
2 min readTerraform vs Vagrant
Terraform for cloud infrastructure vs Vagrant for local VMs. Understand when to use each tool and how they complement each other.
2 min read
CNCF Platform Engineering Technology Radar
The CNCF Q1 2026 Technology Radar focuses on Platform Engineering — 28% of organizations now have dedicated platform teams, Backstage documentary.
4 min read
llm-d: Kubernetes-Native Distributed Inference at Scale
Red Hat's llm-d project enters CNCF Sandbox, bringing disaggregated xPyD serving, tiered prefix caching, and multi-accelerator support to Kubernetes.
3 min read
NVIDIA Goes Platinum at CNCF
NVIDIA joins CNCF Platinum with a $3.8M cloud GPU donation and Kubernetes AI Conformance contributions — marking Kubernetes as the AI infrastructure layer.
3 min read
The AI Inference Challenge: Why Inaction
Jonathan Bryce's CNCF Press Conference keynote reveals the inference gold rush, the maturity paradox of 82% Kubernetes adoption vs 7% daily AI deployment.
3 min read
Kyverno Graduates to CNCF: Policy-Driven AI
Kyverno reaches CNCF Graduation — trusted by Deutsche Telekom, LinkedIn, and Spotify for policy-driven governance of cloud native and AI infrastructure.
3 min read
Huawei Atlas 950 AI SuperPoD: 8,192 NPUs as One Machine
Huawei Atlas 950 SuperPoD connects 8,192 Ascend NPUs via UnifiedBus 2.0 into one machine delivering 8 EFLOPS. Full architecture from chips to openFuyao.
9 min read
openEuler: The Enterprise Linux You Haven't Heard Of
openEuler is a Huawei-originated enterprise Linux with 16 million installations, RISC-V support, and a growing global ecosystem for platform engineers.
6 min read
openFuyao: Open Source Kubernetes Platform
openFuyao is an open source Kubernetes cluster ecosystem with NUMA-aware scheduling, NPU operators, and AI inference acceleration for 10,000+ nodes.
6 min read
AMA Recap: Scaling AI on Kubernetes
Luca Berton joined Packt's Build and Scale AI on Kubernetes workshop for a live AMA on GPU platforms, multi-tenant scheduling, and KubeCon 2026 lessons.
5 min read
Ubuntu 25.10 Installer Broken? Fix the
Ubuntu 25.10 Questing Quokka installer fails when the Subiquity backend stops answering on its Unix socket. Full troubleshooting sequence included.
5 min read
Unsloth on NVIDIA DGX Spark: 2x Faster
Step-by-step guide to running Unsloth on NVIDIA DGX Spark for 2x faster LLM fine-tuning with CUDA 13 and PyTorch containers.
6 min read
What Breaks First When AI Moves from PoC to
The first thing that breaks moving AI from PoC to production on Kubernetes is the assumption that a successful demo equals a production-ready platform.
2 min read
AI Platform First 90 Days on Kubernetes: A Pragmatic Roadmap
Build the first production-ready AI path, then expand. A pragmatic 90-day roadmap for teams starting AI infrastructure on Kubernetes.
4 min read
Autoscale AI Inference on Kubernetes: HPA and KEDA
Inference autoscaling is harder because every extra replica means another GPU and cold-start penalty. Custom metrics like tokens/second are essential.
4 min read
The Hidden Cost Drivers in AI Workloads on Kubernetes
The obvious cost is GPUs. The hidden costs make GPUs sit idle. Three buckets of AI waste — compute, data-path, and operational — and fixes.
4 min read
The Architecture Decisions That Are Hardest
You can replace tools later. Unwinding the wrong tenancy model or GPU scheduling policy is much harder. Get these decisions right early.
4 min read