Skip to content

Savings Calculator

Note: This page is an interactive calculator that requires JavaScript. If you are viewing this on GitHub, visit the live version instead.

Estimate how much you could save by right-sizing your Kubernetes workloads with attune. Enter your current resource allocation and actual usage below, and see the projected monthly and annual savings instantly.

The calculator uses the same pricing model as the operator's built-in EstimatedMonthlySavings computation (configurable via AttuneDefaults).

Quick Presets

Start with a typical scenario, then customize the numbers to match your environment.

Cloud Pricing

AWS on-demand ~$0.031, GCP ~$0.034, Azure ~$0.036
AWS ~$0.004, GCP ~$0.005, Azure ~$0.004

Your Workloads

Enter each service's current resource requests, actual P95 usage, and replica count.

Service name CPU req (m) CPU P95 (m) Mem req (Mi) Mem P95 (Mi) Replicas

Projected Savings with Attune

Monthly savings
$0
Annual savings
$0
Resource reduction
0%
CPU utilization: 0% requested → 0% after right-sizing
Memory utilization: 0% requested → 0% after right-sizing
Service CPU: current → right-sized Memory: current → right-sized Monthly savings
Under-provisioned workloads detected. Rows marked with a warning have P95 usage that exceeds the current resource request (after applying the overhead). Right-sizing these workloads would increase their requests, preventing throttling and OOMKills. This improves reliability rather than reducing cost.
How this is calculated: For each workload, the right-sized value is P95_usage x (1 + overhead/100). Monthly cost uses (cores x CPU_price + GiB x mem_price) x 730 hours. Savings are the difference between current and right-sized costs across all replicas. Actual savings may be higher due to improved bin-packing enabling node consolidation.

Understanding the Numbers

How right-sized values are calculated

For each workload, the right-sized resource request is:

right_sized = P95_usage x (1 + overhead/100)

This matches the operator's default algorithm: take the 95th percentile of observed CPU usage (or 99th for memory), multiply by the overhead, and clamp to the configured bounds.

Why actual savings may be higher

This calculator computes direct resource savings, the difference between what you're requesting now and what you'd request after right-sizing. But the real impact goes further:

  • Node consolidation: Lower pod requests mean better bin-packing. Your cluster autoscaler (or Karpenter) can fit more pods per node and scale down unused nodes. This often adds 10-30% on top of the per-pod savings.

  • Reduced spot/reserved waste: Right-sized workloads let you buy smaller reserved instances or committed use discounts, matching actual need instead of peak overestimate.

  • Operational time saved: No more quarterly resource review meetings. No more "why is the cluster full when utilization is 8%" investigations.

Common findings

Cluster size Typical monthly waste Typical reduction
5-10 services $200-1,000 40-70%
20-50 services $2,000-10,000 50-75%
100+ services $15,000-100,000+ 40-65%

These ranges are based on industry benchmarks from CAST AI, Datadog, and ScaleOps reports.

Ready to capture these savings?