01
blind spot
No Per-Container Visibility
AWS Cost Explorer shows you EC2 bills. Kubecost estimates pod costs. Neither tells you what a single container actually consumes in energy and compute — the real drivers of your bill.
kubmin
by ksctl
The kubmin dispatch Vol. 01 · Cost & Efficiency Deploying fast?
Know what every deploy costs —
in money, energy, and efficiency.
Everyone ships faster with AI. Nobody checks the cost. Kubmin does.
Per-workload cost and efficiency tracking across deployments. The missing layer between your workloads and your bill — now for clusters you create or import.
Kepler (CNCF) /SCI-aligned (GSF) /EKS/AKS/GKE/K3s/Kubeadm /OpenTelemetry
per-workload visibility ·release-vs-release diff ·SCI + SEE scoring ·quick wins, kubectl-ready
The Problem · ch. 01
You're spending thousands on Kubernetes.
But do you know what each workload actually costs?
Three blind spots, every cluster, every release.
02
blind spot
Overprovisioned and Flying Blind
Most teams request 2-4x the resources their workloads actually use. Without per-workload waste detection, you're paying for idle CPU and memory every hour of every day.
03
blind spot
Deploys Make It Worse (Silently)
You ship a new image version — maybe AI-generated, maybe hand-written. CPU goes up 15%. Memory creeps up. Nobody notices because no tool tracks efficiency changes between deployments. The waste compounds with every release.
See It In Action · field test
Three Redis versions.
Three Redis versions.
Same workload.
Wildly different efficiency.
We ran redis-benchmark across versions 8.6.1, 8.4.0, and 8.2.1. Here's what kubmin found — no other tool surfaces this.
fig. 01
63%
more CPU — 8.4.0 vs 8.2.1
(243m vs 149m)
fig. 02
31%
more energy — 8.4.0 vs 8.6.1
(0.0085 vs 0.0064 kWh)
fig. 03
$81
/year wasted — 8.6.1 vs 8.4.0
($0.0297 vs $0.0389/hr median)
fig. 04
31%
more efficient — 8.2.1 vs 8.4.0
on SEE score
8.4.0 uses 87.8% CPU but isn't the most expensive. 8.2.1 uses 51.2% CPU but costs more. No single metric tells the full story — you need cost, energy, and efficiency together. That's what kubmin gives you.
This is one workload. Imagine what kubmin finds across your entire cluster.
Try It On Your Workloads What Kubmin Does · ch. 02
Eight things no other tool does together
The kubmin moat
Each capability ships standalone. Together they form a unified ledger — cost, energy, and efficiency, per workload, per release, across clusters you create or import.
01 Your deploy's real footprint
Per-Image-Version Deep Profile
Point kubmin at a specific image tag and get hourly SCI, energy consumption, CPU, memory, and cost — plus the full trend across that version's lifetime. Not generic pod metrics. Version-scoped visibility into what each build actually costs and consumes in production.
02 Efficiency measured per release
Release-over-Release Efficiency Diff
Compare any two image versions across CPU, memory, energy, SCI, SEE, and runtime cost — per functional unit of work. Ship a new release and see within hours whether it regressed 30% on efficiency, before the regression compounds into your bill.
03 Actually know your workload
Workload Profile Classifier
Kubmin classifies each workload as CPU-bound, memory-bound, general, or resource-starved — automatically, from runtime behavior. Idle-time percentage and hourly heatmaps show exactly when waste spikes. No more right-sizing blind.
04 Use the clusters you already run
Existing Cluster Import
Bring existing EKS, AKS, GKE Standard, or self-managed Kubernetes clusters into Kubmin. Managed imports discover clusters by cloud and region; self-managed imports use kubeconfig. Imported clusters keep their infrastructure ownership while Kubmin adds workload intelligence on top.
05 Bring your stack or create one
Guided Observability
Use your existing Prometheus, Tempo, Loki, and Grafana stack, or let Kubmin deploy the missing pieces. Kubmin configures an OpenTelemetry Collector bridge for metrics, logs, and traces, plus Kepler for energy attribution.
06 Right cluster before the first pod
Smart Cluster Setup and Import
Spinning up a new cluster or importing an existing one? Kubmin filters instance types against your workload profile and shows spec, hourly cost, and embodied carbon for each option. Region picker surfaces real grid carbon intensity and % renewable energy — so cluster decisions are informed, not guessed.
07 Move once, save every hour
Placement & Region Recommender
Running on the default instance type in us-east-1? Kubmin surfaces alternative instance types with similar or better performance at lower cost, plus alternative regions that cut your workload's running cost and run on a cleaner grid. Side-by-side tradeoffs, not guesswork.
08
Quick Wins — Ranked by Savings
Kubmin turns every finding — idle replicas, overprovisioned limits, cheaper instance types, greener regions — into a ranked action list with estimated monthly cost savings and efficiency gains attached. Each win ships with a copy-paste kubectl command. Open the dashboard, act on the top three.
Dollar-ranked actions, not alerts why energy · ch. 03
CPU metrics
lie.
CPU metrics
lie.
Energy
doesn't.
A workload can show 40% CPU utilization and still be wasting energy. How? Polling loops, idle connections, inefficient memory access patterns — they all consume power without producing value. Traditional monitoring tools see 40% busy and move on. Kubmin sees the full picture — CPU usage alongside actual energy consumption over time — so you can spot the workloads where high utilization doesn't mean productive utilization.
Kubmin uses Kepler — a CNCF project — to estimate energy consumption at the container level. Kepler uses kernel-level instrumentation and ML-based models to attribute power usage per container, even in cloud environments where hardware counters aren't directly accessible.
a note on accuracy
In cloud environments, energy measurements are estimates based on ML models trained on real hardware data. They're highly reliable for relative comparison and trend analysis — which is exactly what kubmin needs. The patterns matter more than the absolute numbers.
Get Started · ch. 04 Three steps to see
Three steps to see
what you're wasting
01
step
Create or Import a Cluster
Start with a new ksctl-managed cluster or import an existing EKS, AKS, GKE Standard, or self-managed Kubernetes cluster. Managed imports discover clusters by cloud and region; self-managed imports use kubeconfig.
02
step
Choose Your Observability Path
Connect existing Prometheus, Tempo, Loki, and Grafana endpoints, or let Kubmin deploy the missing pieces. Kubmin always installs the agent, Kepler, Gateway API, and an OpenTelemetry Collector bridge.
03
step
Instrument Apps and See Results
Point your apps at Kubmin's in-cluster OTLP gRPC or HTTP endpoint. Kubmin turns metrics, logs, traces, and Kepler energy data into workload cost, efficiency, SCI/SEE, and Quick Wins.
Start Free
Imported clusters must have cert-manager installed and healthy. Kubmin installs Gateway API CRDs during import and configures the Kubmin agent, Kepler, and OpenTelemetry Collector automatically.
New capability · ch. 04b Guided Observability
Guided Observability
without glue work
Keep your Prometheus, Tempo, Loki, and Grafana stack, or let Kubmin deploy the missing pieces during create/import. Kubmin wires telemetry back to the cost and energy ledger, so metrics, logs, traces, and Kepler data support the same workload decisions.
01
Bring existing Prometheus, Tempo, Loki, and Grafana endpoints
02
Use Kubmin's in-cluster OpenTelemetry Collector for OTLP gRPC and HTTP
03
Let Kubmin deploy missing backends while always installing the agent, Kepler, Gateway API, and collector
For imported clusters, cert-manager must already be installed and healthy. Kubmin installs Gateway API CRDs during import.
The Gap · ch. 05 What your current tools
What your current tools
can't tell you
Kubmin doesn't replace your monitoring stack. It connects cluster onboarding, OpenTelemetry, energy, and workload cost intelligence around it.
| AWS Cost Explorer | Kubecost | Prometheus + Grafana | kubmin | |
|---|---|---|---|---|
| 01 Per-container cost tracking | EC2 only | Est. | Manual | Yes |
| 02 Energy consumption per workload | Yes | |||
| 03 Existing cluster import | Manual | Manual | EKS/AKS/GKE + kubeconfig | |
| 04 Guided observability setup | Manual | Bring yours or deploy | ||
| 05 OpenTelemetry logs/traces/metrics bridge | Manual | OTEL Collector | ||
| 06 GKE Standard import | Manual | Yes | ||
| 07 Idle workload detection | Basic | Manual | Auto | |
| 08 Deployment version comparison | Yes | |||
| 09 SCI + SEE sustainability scoring | Yes | |||
| 10 Regional cost comparison | Partial | Yes | ||
| 11 Ready-to-use kubectl fixes | Yes | |||
| 12 Instance type optimization | Yes | |||
| 13 Setup time | N/A | ~30 min | Hours/Days | Minutes to create/import |
Kubmin doesn't replace your monitoring stack. It sees what your monitoring stack can't — the energy layer that drives real waste, plus the release context that turns metrics into action.
Pricing · ch. 06 Start free.
Start free.
Scale when you need to.
Fixed monthly pricing. No per-node fees. No surprise bills. All plans include the Kubmin agent, Kepler energy insights, and Guided Observability with OpenTelemetry Collector.
Explore
tier 01 $0 /month
- 1 cluster to create or import
- 1 workload
- 1 team member
- Guided Observability setup
- 6h hourly metrics retention
- Basic workload overview
- Community support
most popular
Affordable
tier 02 $19.99 /month
- 5 clusters to create or import
- 5 workloads
- 10 team members
- 72h (3d) hourly metrics retention
- Full workload analysis
- Deployment comparison
- Quick Wins with kubectl commands
- Bring or deploy observability backends
- Priority support
Premium
tier 03 $49.99 /month
- 20 clusters to create or import
- 20 workloads
- 50 team members
- 168h (7d) hourly metrics retention
- Everything in Affordable
- Dedicated support
No hidden fees · Cancel anytime · Bring your stack or let Kubmin deploy one
Built on Proven Foundations
foundation · 01
CNCF Project
Energy measurement powered by Kepler — kernel-level container energy attribution
foundation · 02
OpenTelemetry-native
Kubmin installs an OpenTelemetry Collector bridge for OTLP metrics, logs, and traces while keeping Prometheus, Tempo, Loki, and Grafana optional.
foundation · 03
Green Software Foundation
SCI-aligned sustainability scoring built on the GSF specification
foundation · 04
Open Source
Cluster management powered by ksctl — fully open source on GitHub
finale The bill is running Every hour you wait, your clusters
Every hour you wait, your clusters
keep wasting.
Sign up free. Connect a cluster. See your waste in minutes.
kubmin.ksctl.com · sign in with github · setup in minutes
addendum
Frequently
Frequently
asked.
Honest answers, no marketing fluff. If we don't have the answer here, ask in Discord.