Use these sample values to test the calculator quickly.
| Workload | Window (h) | vCPU | CPU % | Mem (GB) | Mem % | Storage (GB) | Used (GB) | Egress (GB) | $/h |
|---|---|---|---|---|---|---|---|---|---|
| api-gateway-prod | 24 | 8 | 42 | 32 | 55 | 500 | 260 | 120 | 0.85 |
| batch-etl-nightly | 168 | 16 | 18 | 64 | 22 | 1200 | 420 | 35 | 1.60 |
| stream-processor | 24 | 12 | 67 | 48 | 61 | 800 | 620 | 310 | 1.25 |
- CPU_util = CPU_average_used%
- Mem_util = Memory_average_used%
- Storage_util = (Storage_used / Storage_provisioned) × 100
- Weighted_util = (wC·CPU + wM·Mem + wS·Storage) / (wC+wM+wS)
- Compute_cost = hours × $/hour
- Storage_cost = GB × $/GB-month × (hours/730)
- Egress_cost = GB × $/GB
- Idle_share ≈ Idle_cost / Total_cost
- Score = 0.65·Weighted_util + 0.35·(1-Idle_share)·100
This score is a practical estimate, not a provider billing model. Use it to compare scenarios and track improvements over time.
- Pick a time window from your monitoring dashboard.
- Enter allocated resources and average utilization percentages.
- Fill in blended costs for compute, storage, and egress.
- Adjust weights to match your workload’s bottleneck.
- Press Calculate and review score, costs, and recommendations.
- Download CSV or PDF to share in review meetings.
It blends weighted utilization with an estimated idle-cost share. Higher scores usually mean better right-sizing, lower waste, and a healthier spend-to-usage match.
Use averages from your monitoring tool for the same time window. For spiky workloads, consider a weekly window and validate against peak needs before downsizing.
Weights let you reflect what matters most for a workload. For example, memory-bound services can increase memory weight so the score aligns with real constraints.
Storage cost is calculated as provisioned gigabytes times the monthly rate, multiplied by hours/730. This approximates a typical month and keeps comparisons consistent.
Enter the average allocated resources over the window, or a representative steady-state value. Then compare scenarios by adjusting allocations to see how the score shifts.
Egress influences total cost and can raise the idle-cost share if it dominates spend. The recommendations also flag when egress becomes a major cost driver.
Yes. Treat allocated resources as reserved or average provisioned capacity. For containers, use requested resources and average utilization from your cluster metrics.
Start with right-sizing (reduce over-allocated CPU/memory), add schedules for noncritical services, resize storage, and reduce egress through caching or data locality.