Deterministic detector registry
Deterministic, evidence-first detectors across AWS, Azure, GCP, Kubernetes, and AI workloads.Sample catalog entries · Waste figures illustrative onlyPublic Registry (373 JSON) ↓
Identifies Amazon VPC NAT Gateways with near-zero throughput draining $0.056/hr 24/7.
Locates unattached EBS block storage volumes left behind after EC2 instance termination.
Identifies unassociated public IPv4 addresses incurring universal AWS scarcity surcharges.
Exposes invisible incomplete multipart upload chunks quietly accumulating S3 Standard storage fees.
Uncovers dev and staging RDS databases configured with redundant Multi-AZ standbys that double instance fees.
Uncovers log groups configured with 'Never Expire' retention accumulating irreversible ingestion and retention debt.
Detects legacy gp2 EBS volumes paying 20% higher base per-GB rates than modern gp3 baseline volumes.
Identifies running EC2 instances with sustained P95 CPU < 5% and network throughput < 5 KB/s over 14 days.
Detects active Classic ELBs or ALBs with zero healthy registered backend targets incurring continuous hourly fees.
Exposes Redis or Memcached clusters with zero active client connections and near-zero CPU over a 7-day baseline.
Detects DynamoDB tables with < 20% provisioned capacity utilization that should switch to On-Demand.
Identifies manual RDS snapshots retained > 90 days after source database instances were decommissioned.
Locates multi-node OpenSearch/Elasticsearch domains running 24/7 with zero search and indexing queries.
Detects objects < 128 KB transitioned to Standard-IA incurring 128 KB minimum billable capacity penalties.
Finds unattached ENIs left behind after EC2/EKS churn holding billable Elastic IPs or blocking CIDRs.
Correlates multi-day CPU, network, and disk I/O baselines to detect abandoned compute capacity.
Measures CPU and network throughput over 14 days to detect unmanaged deallocated VMs.
Evaluates sustained P95 CPU utilization < 2% on running instances.
DCGM memory telemetry detecting unutilized H100/A100 GPU VRAM reservations.
Compares container resource requests against actual P95 CPU/memory telemetry.
Tracks orphaned managed disks in unattached state across resource groups.
Identifies Google Cloud persistent disks in READY state with zero active VM attachments.
Detects redundant un-cached prompt tokens across autonomous agent execution loops.
Tracks unbound PVC resources retaining underlying storage without active pods.