Workload Optimization
Finds what to right-size, schedule, clean up or move to Spot across VMs, Kubernetes, serverless and AI workloads, and prepares each change.
Workload Optimization is the CloudIQ solution for the instances, clusters, functions and model endpoints behind most of a cloud bill. It finds what to right-size, schedule, clean up or move to Spot, and prepares each change.
What it looks for#
| Workload | What the checks find |
|---|---|
| Virtual machines (EC2, Compute Engine) | Idle instances, older families, Graviton candidates, instances that can stop out of hours |
| Kubernetes (EKS, GKE) | Extended support surcharges, idle node groups, oversized resource requests, idle namespaces |
| Containers and serverless (ECS, Fargate, Lambda, Cloud Run) | Services that can scale to zero, tasks and functions with too much CPU or memory |
| Scheduling | EC2 and RDS that can stop outside working hours, judged on peak as well as average CPU |
| Spot capacity | Auto Scaling groups scored 0 to 100 for Spot suitability |
AI endpoints and accelerators get their own checks:
- SageMaker endpoints that are idle, oversized, or scaled for demand that never came
- Idle notebooks, and failed or repeated training jobs
- Bedrock traffic that could run as batch inference, at about half the price
- Idle provisioned throughput
- Idle GPUs, and GPUs that could be shared between workloads
What you get#
- Rightsizing as a Terraform pull request or a runbook, with rollback steps.
- A schedule with its saving, and the reason any resource was left out.
- Spot advice with a mixed-fleet saving, or a recommendation against Spot where interruptions would cost more than the discount.
- AI endpoint changes such as a lower scaling minimum, batch routing, or a notebook shutdown policy.
How a recommendation is judged#
A rightsizing change built on a quiet week is how a team gets paged. So each finding shows the time window, the usage percentiles, and the peak it allowed for, plus a risk level.
Recommendations from AWS Compute Optimizer and Google Cloud Recommender are pulled in and ranked next to CloudIQ's own, each with its evidence.
AI assistance#
Fini, the CloudIQ cost analyst, can explain any workload finding, look into what changed, and draft the ticket. A scheduled AI assistant looks into cost changes on a cadence you set and sends each engineer a digest of their open tasks. Your engineers approve every change. See Fini and AI assistance.
Next#
Try this against your own spend
A free savings analysis runs the checks described in these docs against your AWS or Google Cloud account and returns a ranked list of priced findings.