Kubernetes cost and performance "what-if" simulator.
Snapshot your cluster, simulate changes, and see the cost impact before making them.

---

😵 The Problem

Kubernetes billing is a black box. Teams overprovision because the cost of an outage outweighs the cost of waste — and nobody has time to model the tradeoffs. Cloud cost tools tell you what you're spending, but they don't answer "what would happen if I changed X?" You're flying blind every time you touch resource requests, node pools, or scheduling policies.

💡 The Solution

KubeWise is a Kubernetes cost × performance "what-if" simulator. It snapshots your live cluster, lets you define hypothetical changes, simulates the Kubernetes scheduler against the modified state, and reports cost savings alongside reliability risk — all without touching your cluster.

The core loop: snapshot → mutate → simulate → compare.

1. 📸 Snapshot the cluster. Read everything that matters: pod resource usage, node capacities, autoscaler configs, PDB rules, affinity constraints, and current cloud pricing. This becomes the baseline with a real dollar cost attached.

2. ✏️ Define a scenario. Describe a hypothetical change:

   • "Right-size every pod's requests to p95 actual usage + 20% buffer"
   • "Move all stateless services to spot instances"
   • "Consolidate from 3 node pools to 2 using larger instance types"
   • "What if traffic doubles next month?"

3. ⚙️ Simulate. Replay the Kubernetes scheduler's bin-packing algorithm against the modified scenario. Run autoscaler logic. Check affinity rules, PDBs, topology constraints. Model spot interruption probability.

4. 📊 Compare and score. Side-by-side: baseline vs. scenario. Cost delta, reliability delta, and confidence intervals. Not a single number — a distribution.

🎯 Differentiator

Existing tools either show you the past (Kubecost) or act on the future without you (CAST AI). Nobody lets you explore the future yourself. KubeWise sits in the gap — high insight AND high user agency.

	Observability tools	Autopilots	KubeWise
Shows cost data	✅	✅	✅
Recommends changes	❌	✅	✅
Simulates before acting	❌	❌	✅
User controls the decision	✅	❌	✅
Runs client-side (no SaaS)	❌	❌	✅

✨ Features

• 📉 Right-size workloads based on actual P50/P90/P95/P99 usage with configurable safety buffers
• 🔗 Simulate node consolidation to find the minimum number of nodes needed
• 💰 Estimate spot savings with eviction risk assessment per workload
• 🤖 CI/CD integration via GitHub Action that posts cost impact as PR comments
• 🔒 Runs entirely client-side — no SaaS dependency, your cluster data never leaves your machine

🚀 Quick Start

Install

# Via krew
kubectl krew install whatif

# Via Go
go install github.com/tochemey/kubewise/cmd/kubectl-whatif@latest

Run

# See current cost breakdown
kubectl whatif snapshot

# Simulate right-sizing
kubectl whatif rightsize --percentile=p95 --buffer=20

# Simulate node consolidation
kubectl whatif consolidate --node-type=m6i.xlarge

# Simulate spot migration
kubectl whatif spot --min-replicas=2 --discount=0.65

Example output

KubeWise: Right-size simulation (p95 + 20% buffer)

  Current monthly cost:    $14230
  Projected monthly cost:  $9840
  Savings:                 $4390/mo (30.8%)  low
  Cluster OOM risk:        0.8%              low

  Top savings by namespace:
    api-gateway          $1200/mo saved    risk: low
    data-pipeline        $980/mo saved     risk: moderate
    ml-inference         $870/mo saved     risk: high
    web-frontend         $640/mo saved     risk: low
    auth-service         $700/mo saved     risk: low

📁 Output Formats

kubectl whatif rightsize                    # Terminal table (default)
kubectl whatif rightsize --output=json      # JSON for scripting
kubectl whatif rightsize --output=markdown  # Markdown for PR comments

📝 Scenario Files

Define reusable scenarios as YAML files:

apiVersion: kubewise.io/v1alpha1
kind: RightSize
metadata:
  name: conservative
  description: "Conservative right-sizing"
spec:
  percentile: p95
  buffer: 30
  scope:
    namespaces: ["*"]
    exclude:
      - namespace: kube-system
  limits:
    strategy: ratio

kubectl whatif apply -f scenario.yaml
kubectl whatif compare -f scenario-a.yaml -f scenario-b.yaml

See docs/scenarios.md for all scenario types and options.

🤖 CI/CD Integration

Add to your GitHub workflow:

- uses: kubewise/action@v1
  with:
    kubeconfig: ${{ secrets.KUBECONFIG }}
    scenario: rightsize
    percentile: p95
    buffer: 20
    comment: true

📚 Documentation

• Quick Start
• Scenarios
• Architecture

🤝 Contributing

make lint        # Run linters
make test-unit   # Run unit tests
make test-all    # Lint + test
make build       # Build binary

📄 License

Apache License 2.0. See LICENSE for details.