/ledger

Optimization Strategies

Purpose: Detailed methods for right-sizing, RI/SP, Spot strategy, and waste elimination. Read when: Concrete cost optimization recommendations are needed.

Right-Sizing Framework

Data Collection Requirements

Instance Family Selection Guide

Right-Sizing Decision Tree

CPU < 10% for 14d?
├── Yes → Is workload bursty?
│   ├── Yes → Switch to burstable (t3/e2/B-series)
│   └── No → Downsize 2 tiers
├── CPU 10-40%?
│   └── Downsize 1 tier, monitor 7d
├── CPU 40-70%?
│   └── Appropriate — no change
└── CPU > 70%?
    ├── Sustained → Scale up 1 tier
    └── Spiking → Add autoscaling

Reserved Instance / Savings Plan Strategy

AWS Commitment Options

GCP Commitment Options

Azure Commitment Options

Coverage Analysis Template

## Commitment Coverage Analysis

### Current State
- Total monthly compute spend: $X,XXX
- On-Demand: $X,XXX (N%)
- Existing commitments: $X,XXX (N%)
- Coverage gap: $X,XXX (N%)

### Recommendation
| Commitment Type | Term | Amount | Expected Savings |
|-----------------|------|--------|-----------------|
| Compute SP | 1yr No Upfront | $X,XXX/hr | $X,XXX/mo (N%) |
| EC2 RI | 1yr Partial | N × instance | $X,XXX/mo (N%) |

### Risk Assessment
- Break-even: N months
- Flexibility impact: [description]
- Rollback plan: [description]

Spot / Preemptible Strategy

Suitability Assessment

Interruption Handling

SPOT_RESILIENCE:
  diversification:
    - Use 4+ instance types per pool
    - Spread across 3+ AZs
    - Mix generations (e.g., m5 + m6i + m7i)

  handling:
    - 2-minute warning handler (AWS)
    - Graceful shutdown hooks
    - Checkpoint/resume for long tasks
    - Automatic fallback to on-demand

  capacity_strategy:
    - lowest-price: Cost-optimized (risk of mass interruption)
    - capacity-optimized: Stability-optimized (recommended)
    - price-capacity-optimized: Balanced (AWS default)

Waste Elimination Checklist