Cluster
Period
🔔4
Executive Summary
prod-us-east-1 · Last 30 Days · All Namespaces
Top-Line KPIs
Total Cluster Spend
$84,320
Idle Resource Waste
$18,740
Efficiency Score
62%
Potential Savings
$23,100
📈 Daily Spend Trend
Total cluster cost vs budget · 30 days
🍩 Cost by Category
Compute / Storage / Network breakdown
🏷️ Top Namespaces by Cost
Monthly spend ranked · all teams
🎯 Resource Utilization
CPU / Memory / Storage avg
Recent Activity
Cost Events — Last 7 Days
8 events| Date | Event | Namespace | Delta | Status |
|---|---|---|---|---|
| Mar 18 | Deployment scaled up (8→24 pods) | payments-svc | +$1,840 | Critical |
| Mar 17 | PVC storage class changed | analytics | +$320 | Warning |
| Mar 16 | Idle pods terminated (rightsizing) | ml-training | -$640 | Saved |
| Mar 15 | Spot instance reclaimed, fallback | batch-jobs | +$210 | Warning |
| Mar 14 | Reserved instance applied | frontend | -$2,100 | Saved |
| Mar 13 | New workload deployed | data-pipelines | +$780 | Info |
Cluster Cost Analysis
Node-level spend · compute breakdown · cloud billing
Node Count
48
Compute Cost
$52,100
On-Demand vs Reserved
64%
Spot Savings
$8,420
🏗️ Cost by Node Pool
Spend distribution across node groups
⏱️ Hourly Cost Heatmap
Average hourly spend by day of week
Node Inventory
Node Details
48 nodes| Node | Instance Type | Pool | CPU Util | Mem Util | Daily Cost | Billing |
|---|
Namespace Analysis
Per-namespace allocation · team attribution · cost trends
📅 Namespace Cost Over Time
Stacked spend — top 5 namespaces · 30d
💡 Request vs Actual Usage
Requested resources vs actual consumption
Resource Efficiency
CPU, memory, GPU utilization · request vs limit gaps
41%
CPU EFFICIENCY
(avg actual / requested)
(avg actual / requested)
58%
MEMORY EFFICIENCY
(avg actual / requested)
(avg actual / requested)
29%
GPU EFFICIENCY
(utilization rate)
(utilization rate)
📉 CPU: Requested vs Used (7d avg)
By workload · millicores
📉 Memory: Requested vs Used (7d avg)
By workload · GiB
Limit vs Request Gaps
Workloads with Oversized Requests
18 workloads| Workload | Namespace | CPU Req | CPU Avg | Mem Req | Mem Avg | Waste/mo | Action |
|---|
Idle Resources
Unused pods · orphaned PVCs · zero-traffic services
Idle Pods
34
Orphaned PVCs
8
Zero-Traffic Svcs
12
Unused ConfigMaps
67
Total Idle Waste
$6,210
💤 Idle Cost Trend
Rolling 30-day idle resource spend
📊 Idle by Namespace
Which teams own the most idle resources
Idle Resource Inventory
All Idle Resources
34 items| Resource | Type | Namespace | Idle Since | CPU (unused) | Mem (unused) | Cost/mo | Sev |
|---|
Rightsizing Recommendations
AI-powered resource right-sizing · VPA-ready suggestions
Total Savings if Applied
$12,840
Workloads Analyzed
142
Overprovisioned
68
Underprovisioned
9
🏆 Top Rightsizing Opportunities
Estimated monthly savings by workload
🔵 Provision State Distribution
Over / under / correctly provisioned
Recommendations
Rightsizing Queue
68 items| Workload | Namespace | Current CPU | Rec CPU | Current Mem | Rec Mem | Savings/mo | Confidence |
|---|
Workload Breakdown
Per-pod, per-deployment cost attribution
🚀 Cost by Deployment
Top 8 most expensive
📦 Cost by Controller Type
Deployment / DaemonSet / Job / CronJob
📈 Pod Count vs Cost
Correlation over 30 days
All Workloads
Workload Cost Table
89 workloads| Workload | Namespace | Type | Pods | CPU Cost | Mem Cost | Total/mo | Efficiency |
|---|
Network Cost Analysis
Egress · cross-zone · inter-namespace traffic costs
Total Network Cost
$9,840
External Egress
$5,120
Cross-AZ Traffic
$3,400
Intra-Cluster
$1,320
🌐 Network Cost Breakdown
Egress / cross-AZ / intra-cluster · 30d trend
🗺️ Top Egress Sources
By service generating most outbound traffic
Storage Cost Analysis
PVC usage · storage classes · snapshot costs
Total Storage Cost
$12,380
Total PVC Capacity
48 TB
Avg Utilization
54%
Orphaned PVCs
8
💾 Cost by Storage Class
gp3 / io1 / standard / ephemeral
📊 PVC Usage vs Provisioned
Top 10 PVCs — actual vs claimed capacity
Chargeback / Team Attribution
Cost allocation by team · showback reports · budget tracking
Teams Tracked
9
Over Budget
2
Tagged Resources
91%
Unallocated Overhead
$6,240
🏢 Team Spend vs Budget
Actual vs allocated budget · all teams
📅 Monthly Chargeback Trend
Team spend history · 6 months
Team Budget Status
Chargeback Summary
| Team | Namespace(s) | Budget | Actual | Variance | % Used | Status |
|---|
Alerts & Anomalies
Cost spikes · quota breaches · anomaly detection
Critical Alerts
4
Warnings
11
Anomalies (ML)
3
Resolved (7d)
18
Active Alerts
🔴
Cost Spike — payments-svc
Spend increased 312% vs 7-day rolling avg. 24 pods running vs typical 8.
Root cause: load test not terminated. Cluster autoscaler added 6 nodes.
🔴
GPU Idle — ml-training namespace
4x NVIDIA T4 GPUs idle for 31 hours. Training job completed but pods not terminated.
$96/hour wasted on idle GPU nodes.
🟡
Budget Threshold — analytics team at 87%
analytics namespace consumed $20,880 of $24,000 monthly budget.
At current rate, will exceed budget in ~6 days.
🟡
Orphaned PVCs Detected
3 PVCs unattached for >7 days in staging namespace. gp3 storage costing $420/mo.
Likely residue from deleted deployments.
🔵
Anomaly Detected — egress spike
data-pipelines namespace generated 3.2TB egress in 4h (10x baseline).
ML model detected statistical anomaly. Review data export jobs.
Alert History
📊 Alert Volume — 30 Days
Critical / warning / anomaly counts per day
Savings Roadmap
Prioritized cost reduction opportunities · implementation timeline
Quick Wins (this week)
$6,420
This Month
$10,840
This Quarter
$23,100
Annual Impact
$277K
Top Opportunities
🎯
Terminate Idle GPU Pods
4 GPU pods in ml-training idle 31h. Implement job completion hooks or auto-terminate after TTL.
$3,456/mo
⚡ Quick win · 1h effort
⬇️
Rightsize Overprovisioned Workloads
68 workloads using <40% of requested CPU/memory. Apply VPA recommendations progressively.
$12,840/mo
📋 Medium effort · 1 week
🏷️
Purchase Reserved Instances
64% on-demand exposure on baseline workloads. 1yr RI commitment saves ~38% on stable nodes.
$8,200/mo
📋 Low effort · Billing change
🌐
Reduce Cross-AZ Traffic
Topology-aware routing can reduce cross-AZ calls. Add pod affinity rules to co-locate chat services.
$3,400/mo
🔧 Medium effort · 2 weeks
💾
Clean Orphaned PVCs
8 unattached volumes costing $1,120/mo. Implement PVC lifecycle policy + automated cleanup job.
$1,120/mo
⚡ Quick win · 2h effort
☁️
Migrate to Spot for Batch Jobs
batch-jobs namespace runs on on-demand. Batch/fault-tolerant workloads ideal for spot (70% discount).
$4,200/mo
🔧 Medium effort · 3 weeks
Implementation Timeline
Week 1 — Immediate
Terminate Idle GPU Pods + Clean Orphaned PVCs
No-risk removals. Set up auto-TTL on training jobs. Run PVC audit script. Expected save: $4,576/mo.
Weeks 2–3 — Short Term
Begin Rightsizing Rollout (Wave 1 — dev namespaces)
Apply VPA recommendations to non-prod first. Validate stability. Expected save: $3,200/mo from dev alone.
Month 1 — Medium Term
Rightsizing Wave 2 (prod) + Spot Migration for Batch
Roll rightsizing to prod with canary. Migrate batch-jobs to spot node group. Expected save: $13,640/mo.
Month 2 — Strategic
Reserved Instance Purchase + Topology-Aware Routing
Commit to 1yr RIs for stable baseline capacity. Deploy topology hints in service mesh. Expected save: $11,600/mo.
Quarter End — Ongoing
FinOps Culture: Budget Alerts + Team Dashboards
Embed cost metrics in CI/CD. Teams own their spend. Quarterly review cadence established.