<div id="readability-page-1" class="page"><div><img src="https://devstarsj.github.io/assets/https://images.unsplash.com/photo-1554224155-8d04cb21cd6c?w=1920&amp;q=80">Kubernetes makes scaling easy. Too easy. Teams often overprovision by 3-5x without realizing it. This guide shows you how to find and eliminate waste.<img src="https://images.unsplash.com/photo-1554224155-8d04cb21cd6c?w=800&amp;q=80" alt="Cost analysis"> Photo by <a href="https://unsplash.com/@kmuza">Carlos Muza</a> on Unsplash<h2 id="the-typical-kubernetes-waste-pattern">The Typical Kubernetes Waste Pattern</h2>Most clusters look like this:<div><pre><code>Requested CPU: 1000 cores
Actually Used: 200 cores ← 80% waste

Requested RAM: 2000 GB
Actually Used: 600 GB ← 70% waste
</code></pre></div>Why? Developers request resources based on fear, not data.<h2 id="step-1-measure-everything">Step 1: Measure Everything</h2>You can’t optimize what you don’t measure. Install cost visibility:<h3 id="option-a-kubecost-open-source">Option A: Kubecost (Open Source)</h3><div><pre><code>helm install kubecost cost-analyzer \
 --repo https://kubecost.github.io/cost-analyzer/ \
 --namespace kubecost --create-namespace
</code></pre></div><h3 id="option-b-opencost-cncf">Option B: OpenCost (CNCF)</h3><div><pre><code>helm install opencost opencost \
 --repo https://opencost.github.io/opencost-helm-chart \
 --namespace opencost --create-namespace
</code></pre></div><img src="https://images.unsplash.com/photo-1551288049-bebda4e38f71?w=800&amp;q=80" alt="Dashboard"> Photo by <a href="https://unsplash.com/@lukechesser">Luke Chesser</a> on Unsplash<h2 id="step-2-right-size-requests-and-limits">Step 2: Right-Size Requests and Limits</h2><h3 id="find-over-provisioned-workloads">Find Over-Provisioned Workloads</h3><div><pre><code># Using kubectl-view-allocations plugin
kubectl view-allocations -u

# Or query Prometheus directly
# CPU: requested vs used
sum(kube_pod_container_resource_requests{resource="cpu"}) by (namespace)
/
sum(rate(container_cpu_usage_seconds_total[5m])) by (namespace)
</code></pre></div><h3 id="apply-recommendations">Apply Recommendations</h3><div><pre><code># Before: Developer's guess
resources:
 requests:
 cpu: "2"
 memory: "4Gi"
 limits:
 cpu: "4"
 memory: "8Gi"

# After: Based on P95 usage + 20% buffer
resources:
 requests:
 cpu: "200m"
 memory: "512Mi"
 limits:
 cpu: "500m"
 memory: "1Gi"
</code></pre></div><h3 id="automate-with-vpa">Automate with VPA</h3>Vertical Pod Autoscaler adjusts resources automatically:<div><pre><code>apiVersion: autoscaling.k8s.io/v1
kind: VerticalPodAutoscaler
metadata:
 name: api-vpa
spec:
 targetRef:
 apiVersion: apps/v1
 kind: Deployment
 name: api
 updatePolicy:
 updateMode: "Auto" # or "Off" for recommendations only
 resourcePolicy:
 containerPolicies:
 - containerName: "*"
 minAllowed:
 cpu: "50m"
 memory: "64Mi"
 maxAllowed:
 cpu: "2"
 memory: "4Gi"
</code></pre></div><h2 id="step-3-use-spotpreemptible-nodes">Step 3: Use Spot/Preemptible Nodes</h2>Spot instances cost 60-90% less. Use them for:<ul><li>Stateless workloads</li><li>Batch jobs</li><li>Dev/staging environments</li></ul><h3 id="node-pool-setup-gke">Node Pool Setup (GKE)</h3><div><pre><code>apiVersion: container.google.com/v1
kind: NodePool
spec:
 config:
 spot: true
 autoscaling:
 enabled: true
 minNodeCount: 0
 maxNodeCount: 10
</code></pre></div><h3 id="workload-tolerations">Workload Tolerations</h3><div><pre><code>spec:
 tolerations:
 - key: "kubernetes.azure.com/scalesetpriority"
 operator: "Equal"
 value: "spot"
 effect: "NoSchedule"
 affinity:
 nodeAffinity:
 preferredDuringSchedulingIgnoredDuringExecution:
 - weight: 100
 preference:
 matchExpressions:
 - key: "node.kubernetes.io/lifecycle"
 operator: "In"
 values: ["spot"]
</code></pre></div><h2 id="step-4-scale-down-non-production">Step 4: Scale Down Non-Production</h2>Dev and staging clusters don’t need to run 24/7.<h3 id="kube-downscaler">Kube-downscaler</h3><div><pre><code>apiVersion: apps/v1
kind: Deployment
metadata:
 annotations:
 # Scale to 0 outside business hours
 downscaler/downtime: "Mon-Fri 20:00-08:00 UTC, Sat-Sun 00:00-24:00 UTC"
</code></pre></div><h3 id="cluster-auto-sleep">Cluster Auto-Sleep</h3><div><pre><code># Scale all deployments to 0 replicas
kubectl get deploy -A -o name | xargs -I {} kubectl scale {} --replicas=0

# Or use a CronJob
apiVersion: batch/v1
kind: CronJob
metadata:
 name: scale-down-dev
spec:
 schedule: "0 20 * * 1-5" # 8 PM weekdays
 jobTemplate:
 spec:
 template:
 spec:
 containers:
 - name: kubectl
 image: bitnami/kubectl
 command:
 - /bin/sh
 - -c
 - kubectl scale deploy --all --replicas=0 -n dev
</code></pre></div><h2 id="step-5-optimize-storage">Step 5: Optimize Storage</h2><h3 id="delete-orphaned-pvcs">Delete Orphaned PVCs</h3><div><pre><code># Find unbound PVCs
kubectl get pvc -A | grep -v Bound

# Find PVs without claims
kubectl get pv | grep Released
</code></pre></div><h3 id="use-appropriate-storage-classes">Use Appropriate Storage Classes</h3><div><pre><code># Don't use SSD for logs
kind: PersistentVolumeClaim
spec:
 storageClassName: standard # Not premium-ssd
 resources:
 requests:
 storage: 100Gi
</code></pre></div><h3 id="enable-storage-auto-expansion">Enable Storage Auto-Expansion</h3><div><pre><code>allowVolumeExpansion: true # In StorageClass
</code></pre></div><h2 id="step-6-network-cost-reduction">Step 6: Network Cost Reduction</h2>Cross-AZ traffic is expensive. Keep pods close:<div><pre><code>spec:
 affinity:
 podAffinity:
 preferredDuringSchedulingIgnoredDuringExecution:
 - weight: 100
 podAffinityTerm:
 labelSelector:
 matchLabels:
 app: database
 topologyKey: topology.kubernetes.io/zone
</code></pre></div><h2 id="cost-monitoring-queries">Cost Monitoring Queries</h2><h3 id="prometheusgrafana-alerts">Prometheus/Grafana Alerts</h3><div><pre><code># Alert if namespace cost exceeds budget
- alert: NamespaceCostOverBudget
 expr: |
 sum(
 container_memory_working_set_bytes{namespace="production"} 
 * on(node) group_left() node_ram_hourly_cost
 ) &gt; 1000 # $1000/hour threshold
 for: 1h
 labels:
 severity: warning
</code></pre></div><h3 id="daily-cost-report">Daily Cost Report</h3><div><pre><code>-- Kubecost SQL query
SELECT 
 namespace,
 SUM(cpu_cost + ram_cost + pv_cost + network_cost) as total_cost
FROM allocations
WHERE window_start &gt;= NOW() - INTERVAL '24 hours'
GROUP BY namespace
ORDER BY total_cost DESC
LIMIT 10;
</code></pre></div><h2 id="quick-wins-summary">Quick Wins Summary</h2><table><thead><tr><th>Action</th><th>Effort</th><th>Savings</th></tr></thead><tbody><tr><td>Right-size requests</td><td>Medium</td><td>20-40%</td></tr><tr><td>Spot nodes for stateless</td><td>Low</td><td>30-60%</td></tr><tr><td>Scale down non-prod</td><td>Low</td><td>50-70%</td></tr><tr><td>Delete orphaned resources</td><td>Low</td><td>5-10%</td></tr><tr><td>Pod topology awareness</td><td>Medium</td><td>10-20%</td></tr></tbody></table><table><thead><tr><th>Tool</th><th>Type</th><th>Best For</th></tr></thead><tbody><tr><td>Kubecost</td><td>Full platform</td><td>Enterprise visibility</td></tr><tr><td>OpenCost</td><td>Open source</td><td>Cost allocation</td></tr><tr><td>Goldilocks</td><td>VPA helper</td><td>Right-sizing</td></tr><tr><td>kube-downscaler</td><td>Scheduler</td><td>Non-prod savings</td></tr><tr><td>CAST AI</td><td>Automation</td><td>Hands-off optimization</td></tr></tbody></table><h2 id="action-plan">Action Plan</h2><ol><li>Week 1: Install Kubecost/OpenCost, get baseline</li><li>Week 2: Apply VPA recommendations to top 10 workloads</li><li>Week 3: Add spot nodes, migrate stateless workloads</li><li>Week 4: Set up non-prod downscaling</li><li>Ongoing: Monthly cost reviews, budget alerts</li></ol><hr>The goal isn’t minimum cost—it’s efficient cost. Pay for what you use, and use what you pay for. <div>이 글이 도움이 되셨다면 공감 및 광고 클릭을 부탁드립니다 :) </div></div></div>

Dev Note

Kubernetes makes scaling easy. Too easy. Teams often overprovision by 3-5x without realizing it. This guide shows you how to find and eliminate waste.
Photo by Carlos Muza on Unsplash
The Typical Kubernetes Waste Pattern
Most clusters look like this:
Requested CPU:  1000 cores
Actually Used:   200 cores  ← 80% waste
Requested RAM:  2000 GB
Actually Used:   600 GB    ← 70% waste
Why? Developers request resources based on fear, not data.
Step 1: Measure Everything
You can’t optimize what you don’t measure. Install cost visibility:
Option A: Kubecost (Open Source)
helm install kubecost cost-analyzer \
  --repo https://kubecost.github.io/cost-analyzer/ \
  --namespace kubecost --create-namespace
Option B: OpenCost (CNCF)
helm install opencost opencost \
  --repo https://opencost.github.io/opencost-helm-chart \
  --namespace opencost --create-namespace
Photo by Luke Chesser on Unsplash
Step 2: Right-Size Requests and Limits
Find Over-Provisioned Workloads
# Using kubectl-view-allocations plugin
kubectl view-allocations -u
# Or query Prometheus directly
# CPU: requested vs used
sum(kube_pod_container_resource_requests{resource="cpu"}) by (namespace)
/
sum(rate(container_cpu_usage_seconds_total[5m])) by (namespace)
Apply Recommendations
# Before: Developer's guess
resources:
  requests:
    cpu: "2"
    memory: "4Gi"
  limits:
    cpu: "4"
    memory: "8Gi"
# After: Based on P95 usage + 20% buffer
resources:
  requests:
    cpu: "200m"
    memory: "512Mi"
  limits:
    cpu: "500m"
    memory: "1Gi"
Automate with VPA
Vertical Pod Autoscaler adjusts resources automatically:
apiVersion: autoscaling.k8s.io/v1
kind: VerticalPodAutoscaler
metadata:
  name: api-vpa
spec:
  targetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: api
  updatePolicy:
    updateMode: "Auto"  # or "Off" for recommendations only
  resourcePolicy:
    containerPolicies:
    - containerName: "*"
      minAllowed:
        cpu: "50m"
        memory: "64Mi"
      maxAllowed:
        cpu: "2"
        memory: "4Gi"
Step 3: Use Spot/Preemptible Nodes
Spot instances cost 60-90% less. Use them for:
Stateless workloads
Batch jobs
Dev/staging environments
Node Pool Setup (GKE)
apiVersion: container.google.com/v1
kind: NodePool
spec:
  config:
    spot: true
  autoscaling:
    enabled: true
    minNodeCount: 0
    maxNodeCount: 10
Workload Tolerations
spec:
  tolerations:
  - key: "kubernetes.azure.com/scalesetpriority"
    operator: "Equal"
    value: "spot"
    effect: "NoSchedule"
  affinity:
    nodeAffinity:
      preferredDuringSchedulingIgnoredDuringExecution:
      - weight: 100
        preference:
          matchExpressions:
          - key: "node.kubernetes.io/lifecycle"
            operator: "In"
            values: ["spot"]
Step 4: Scale Down Non-Production
Dev and staging clusters don’t need to run 24/7.
Kube-downscaler
apiVersion: apps/v1
kind: Deployment
metadata:
  annotations:
    # Scale to 0 outside business hours
    downscaler/downtime: "Mon-Fri 20:00-08:00 UTC, Sat-Sun 00:00-24:00 UTC"
Cluster Auto-Sleep
# Scale all deployments to 0 replicas
kubectl get deploy -A -o name | xargs -I {} kubectl scale {} --replicas=0
# Or use a CronJob
apiVersion: batch/v1
kind: CronJob
metadata:
  name: scale-down-dev
spec:
  schedule: "0 20 * * 1-5"  # 8 PM weekdays
  jobTemplate:
    spec:
      template:
        spec:
          containers:
          - name: kubectl
            image: bitnami/kubectl
            command:
            - /bin/sh
            - -c
            - kubectl scale deploy --all --replicas=0 -n dev
Step 5: Optimize Storage
Delete Orphaned PVCs
# Find unbound PVCs
kubectl get pvc -A | grep -v Bound
# Find PVs without claims
kubectl get pv | grep Released
Use Appropriate Storage Classes
# Don't use SSD for logs
kind: PersistentVolumeClaim
spec:
  storageClassName: standard  # Not premium-ssd
  resources:
    requests:
      storage: 100Gi
Enable Storage Auto-Expansion
allowVolumeExpansion: true  # In StorageClass
Step 6: Network Cost Reduction
Cross-AZ traffic is expensive. Keep pods close:
spec:
  affinity:
    podAffinity:
      preferredDuringSchedulingIgnoredDuringExecution:
      - weight: 100
        podAffinityTerm:
          labelSelector:
            matchLabels:
              app: database
          topologyKey: topology.kubernetes.io/zone
Cost Monitoring Queries
Prometheus/Grafana Alerts
# Alert if namespace cost exceeds budget
- alert: NamespaceCostOverBudget
  expr: |
    sum(
      container_memory_working_set_bytes{namespace="production"} 
      * on(node) group_left() node_ram_hourly_cost
    ) > 1000  # $1000/hour threshold
  for: 1h
  labels:
    severity: warning
Daily Cost Report
-- Kubecost SQL query
SELECT 
  namespace,
  SUM(cpu_cost + ram_cost + pv_cost + network_cost) as total_cost
FROM allocations
WHERE window_start >= NOW() - INTERVAL '24 hours'
GROUP BY namespace
ORDER BY total_cost DESC
LIMIT 10;
Quick Wins Summary
Action
      Effort
      Savings
    
Right-size requests
      Medium
      20-40%
    
Spot nodes for stateless
      Low
      30-60%
    
Scale down non-prod
      Low
      50-70%
    
Delete orphaned resources
      Low
      5-10%
    
Pod topology awareness
      Medium
      10-20%
    
Tools Comparison
Tool
      Type
      Best For
    
Kubecost
      Full platform
      Enterprise visibility
    
OpenCost
      Open source
      Cost allocation
    
Goldilocks
      VPA helper
      Right-sizing
    
kube-downscaler
      Scheduler
      Non-prod savings
    
CAST AI
      Automation
      Hands-off optimization
    
Action Plan
Week 1: Install Kubecost/OpenCost, get baseline
Week 2: Apply VPA recommendations to top 10 workloads
Week 3: Add spot nodes, migrate stateless workloads
Week 4: Set up non-prod downscaling
Ongoing: Monthly cost reviews, budget alerts
The goal isn’t minimum cost—it’s efficient cost. Pay for what you use, and use what you pay for.

Kubernetes Cost Optimization: A FinOps Engineering Guide

댓글