07
Demo 07 — Right-Sizing & Cost Impact Analysis
Python 3 (stdlib)
⏱ ~3 min
No cluster needed
Run this demo
View source on GitHub ↗
cd demo-07-rightsizing
chmod +x demo.sh
./demo.shPure Python analysis — no containers, no cluster needed. 14 days of bundled Prometheus sample data from a 7-service cluster produces right-sizing recommendations and a quantified cost business case.
Run options
./demo.sh # bundled sample data
./demo.sh --live # try kubectl first
python3 analyze.py --cost-per-node-hour 0.768 # custom node cost
Sample results
| Service | CPU before -> after | Memory before -> after |
|---|---|---|
| payment-service (SB) | 2000m -> 560m (-72%) | 4096 -> 2304Mi (-44%) |
| fraud-detection (QK) | 1500m -> 280m (-81%) | 2048 -> 880Mi (-57%) |
| report-generator (SB) | 4000m -> 3640m (-9%) | 8192 -> 7744Mi (-5%) |
The
report-generatoris the honest exception — batch workload with real CPU and memory near limit. Not everything gets cut.
Business case output
4 nodes -> 2 nodes · +67% pod density · $6,720/month saving · 17x ROI
OpenShift Cost Management
Console -> Cost Management -> Optimizations
API: GET /api/cost-management/v1/recommendations/openshift/