Waste Detection

Find and remove waste.

Overview

Waste detection is the discipline of routinely finding and removing cloud resources nobody is using. The accumulation is steady: experiments left running, migrations that left old infrastructure behind, services that got rewritten without the originals being decommissioned. Without a recurring sweep, this drift is invisible at the line-item level and significant at the bill level. The discipline is automation plus weekly review, not a heroic quarterly cleanup.

• Find and remove waste. Per-account scan for unused resources; the cost is in the accumulation, not in any single instance.
• Idle EC2 instances. Sub-5 percent CPU sustained for 14 days; almost always experiments or forgotten test infrastructure.
• Unattached volumes and unused EIPs. Per-account orphan scan; volumes detached from terminated instances, EIPs allocated but not associated.
• Old snapshots plus idle load balancers. Snapshots older than retention policy; load balancers with zero requests in the last week; both accumulate silently.

The approach

The practical approach is automated detection running daily, weekly review by the FinOps owner, owner notification before deletion (not after), per-quarter bulk cleanup of items that survived owner review, and documented per-resource-type cleanup policy so the rules are predictable. The discipline is in the cadence; ad-hoc cleanups never keep up with accumulation.

• Automated detection. Daily scan against waste rules; AWS Trusted Advisor, AWS Config, third-party FinOps tooling, or in-house scripts all work.
• Weekly review. FinOps owner reviews the new candidates; the cadence keeps drift bounded between major cleanups.
• Owner notification. Per-resource the owner gets notice before deletion; gives them a window to confirm or reclaim.
• Per-quarter cleanup plus documented policy. Quarterly bulk delete for items that survived owner review; per-resource-type cleanup rule documented for the team handbook.

Why this compounds

Waste detection compounds across accounts and quarters. Each cleanup that sticks reduces baseline cost; each weekly sweep prevents accumulation; the team’s vocabulary for cost hygiene grows. After two quarters, the cleanup is not a project, it is a 30-minute weekly review and the bill stops surprising anyone.

• Cost efficiency. Right resources match consumption; the bill reflects actual usage rather than archeological layers.
• Operational hygiene. Per-account cleanup keeps the resource graph close to the architecture diagram; investigation gets faster because there is less noise.
• Security posture. Fewer unused resources mean fewer forgotten attack surfaces; old EIPs and idle instances are favourite footholds.
• Institutional knowledge. Each cleanup teaches resource lifecycle patterns; the team learns where waste accumulates and prevents it at creation time.

Waste detection is an operational discipline that pays off across years. Nova AI Ops integrates with cost telemetry, surfaces waste patterns, and supports the team’s cost discipline.