WebCost & UsageOptimization

Optimization Views

The Optimization section at /cost-and-usage/optimization provides four analytical views for identifying cost reduction opportunities and usage anomalies. Each view addresses a specific dimension of cloud spending efficiency.

Indicators View

The Indicators view presents three gauge cards that summarize the most critical efficiency signals at a glance:

IndicatorCalculationGood Range
Idle Spend %Cost attributed to resources with zero usage in the period< 5%
OverprovisioningResources with average CPU or memory utilization below threshold (Compute Optimizer signal)< 15%
Usage ConcentrationPercentage of total cost from the top 10 resources (Pareto measure)60–80%

Baseline View (Anomaly Detection)

The Baseline view renders a Z-score anomaly detection chart using a 14-day rolling window:

  • Baseline band: Calculated from days D-16 to D-3 (excluding the most recent 2 days to avoid recency bias)
  • Expected range: Mean ± (2 × standard deviation) drawn as a shaded band
  • Anomaly markers: Points outside the band are flagged in red. Z-score threshold is configurable (default: 2.0)
  • Dimension filtering: Same dimension selector as Unit Cost — pivot by service, region, resource, or tag

Stability View

The Stability view is a scatter plot that maps each resource or service grouping onto a two-axis space:

  • X-axis: Coefficient of Variation (CV = std dev / mean) — measures usage predictability. Lower CV means more stable, predictable usage.
  • Y-axis: Cost magnitude (total spend in period)
  • Commitment zone: Resources in the low-CV, high-cost quadrant are ideal candidates for Reserved Instances or Savings Plans

Hovering a dot shows the resource name, CV value, and estimated annual savings from a 1-year commitment.

Advanced View

The Advanced view consolidates four specialized visualizations in a single scrollable page:

ComponentDescription
Calendar HeatmapMonth-view heatmap coloring each day by spend. Idle days (zero spend) shown in grey. Helps spot weekday vs. weekend patterns.
Zero-Query DaysCount of days in the period where Redshift received zero queries. Indicates cluster idle time — a prime candidate for scheduled pausing.
Off-Hours DonutPercentage of total spend occurring outside business hours (configurable: default 6 PM–8 AM and weekends). High off-hours spend may indicate runaway jobs.
Underutilized ClustersTable of Redshift clusters with average query time <5 minutes and <10 queries/day in the period — candidates for downsizing.

Configuration

All optimization thresholds (idle spend threshold, Z-score cutoff, business hours definition, CV bucket boundaries) are configurable via /cost-and-usage/config. Changes apply to all four views immediately on save.