672 posts tagged with "PaaS"
Platform-as-a-Service tooling, economics, and self-hosted alternatives
Kubernetes 1.36 Can Scale to Zero — But Not Like Render Does: What a Self-Hosted PaaS Still Has to Wire
Kubernetes 1.36 keeps HPA scale-to-zero in alpha, disabled by default and limited to External/Object metrics. A concrete walkthrough of what a Cluster API fleet still needs — KEDA HTTP add-on or Knative activator, external metrics, and probe math — to give tenants real scale-to-zero without renting Render's cold-start model.
Kubero Nailed Kubernetes-Native PaaS — But Somebody Still Has to Build the Cluster
Kubero gives you Heroku-like deploys on Kubernetes with GitOps review apps and 160+ templates, but it won't provision the cluster. A table-driven comparison with Coolify and Dokploy shows where single-box ends, where Kubero fits, and why a Cluster API fleet is the missing machine layer.
Netcup's €8.74 Box vs Hetzner's €42.99 CCX: What a Second EU Provider Actually Costs Your Fleet
After Hetzner's third 2026 hike tripled dedicated CCX pricing, Netcup's RS 1000 G12 offers 4 dedicated EPYC cores for €8.74 — but no CAPH provider, fewer regions, and no declarative machine lifecycle. A fleet-scale recompute with honest gaps.
OVHcloud's Six-Day Outage: What a Multi-Day Recovery Window Actually Costs a Tenant
OVHcloud's web hosting incident took 5 days 18 hours to fully mitigate in H1 2026 — the longest web hosting window IncidentHub tracked in six months. A worked accounting of what 138 hours costs a tenant with no failover, why SLA credits don't cover it, and what the same failure looks like when your fleet remediates itself in minutes.
100 Platforms, One Catch: What Free Hosting Actually Gives You in 2026
A 150-platform free-tier census went viral in 2026. Here's what Vercel, Netlify, Heroku, AWS and a raw VPS actually give you for free — what breaks at month 13 — and what the same hobby stack costs on a €4.49 Hetzner box that never had a free tier to expire.
Your Rented Control Plane Costs More Than Your Servers: What Qovery's $299–$2,099 Agentic Platform Fee Actually Buys
Qovery's $299–$2,099/month control-plane fee sits on top of your cloud bill — not instead of it. A line-by-line accounting of what the agentic platform buys, why agents don't consume seats but do consume deployment minutes, and when owning the fleet costs less than renting the plane that manages it.
Railway Prices Compute in Scientific Notation: What $0.000463 per vCPU-Second Actually Costs a Month
Railway's $0.000463/vCPU-minute is $20 per vCPU per month — but only if you catch the unit. A worked table converts Railway's per-second and per-minute rates to monthly costs for four real workloads, shows why legibility is its own billing risk, and prices the same stack on a flat Hetzner box.
Render's $1.5B Valuation and a 20× Bandwidth Cut in the Same Quarter: Who $0.15/GB Egress Is Actually Priced For
Render raised $100M at a $1.5B valuation in the same quarter it cut Hobby bandwidth 20× — from 100GB to 5GB at $0.15/GB overage. A worked table prices 5GB to 5TB on Render Hobby/Pro/Scale against a Hetzner CX22's flat 20TB to show who the new meter is actually priced for.
51 Days to Fix Object Storage: What a Seven-Week Remediation Tells You About the Real Reliability Bar
Scaleway's AMS object storage incident took 33 to 51 days to fully remediate region by region in H1 2026 — no data loss, but seven weeks where the managed SLA was the thing that was broken.