Rightsizing Lab

Right-sized instances, backed by real performance data.

Rightsizing Lab recommends the instance, node or database size that fits how a workload actually runs. It shows CPU, memory, disk and network for the current size and projects the same load onto the recommended one, so nobody has to guess.

Rightsizing Laborders-db · PostgreSQL · Multi-AZ + 1 replica
CPU now (16 vCPU) Projected on 8 vCPU14 days
100%80% headroom limit0%
Currentdb.r6g.4xlarge16 vCPU · 128 GiB$5,040 / mo
Recommendeddb.r6g.2xlarge8 vCPU · 64 GiB$2,520 / mo
CPU p95 after42%
Memory p95 after34%
IOPS p952,180 / 12,000
Saving$2,520 / mo

What it does

Built for the engineers who make the change, and the finance team that checks it.

Headroom you set

Choose a target such as p95 CPU under 60 percent. Recommendations never cross the limits you define.

Family and generation moves

Suggestions include moves to newer generations and Graviton or AMD families where the workload is compatible.

Kubernetes requests

Pod CPU and memory requests compared with actual use, and node pool shapes that pack workloads more tightly.

Managed databases

RDS, Azure SQL and Cloud SQL sizing based on CPU, memory, IOPS and connection counts.

Projected performance

A chart of the same 14 to 30 days replayed on the recommended size, with peaks highlighted.

Ready for change review

Each recommendation exports as a ticket with the evidence attached for your change process.

How it works

Rightsizing Lab, step by step.

Good to know

  • Recommendations respect your headroom limits
  • Peaks and seasonality are part of every check
  • Savings use your negotiated pricing and discounts
  • Nothing is resized automatically
  1. 01

    Metrics are gathered

    CPU, memory, disk and network metrics are read for every candidate. Memory needs the agent you already run, such as the CloudWatch agent.

  2. 02

    Sizes are compared

    The current shape is compared with every viable size and family in the same region, using your pricing.

  3. 03

    Load is projected

    Historical load is replayed on the candidate size to check that peaks stay within your headroom rules.

  4. 04

    Change is planned

    You get the target size, expected saving and a rollback note. Your team schedules the change.

Questions

About Rightsizing Lab

What if memory metrics are missing?

We mark the recommendation as CPU-only and lower its confidence. You will see exactly which metric is missing and how to enable it.

Do you consider reservations already bought?

Yes. A resize that would leave an existing reservation unused is flagged, so you can time the change with the commitment.

Your forecast

Find your savings free.

Each month we spendonand we want next month to be clearer.

Read-only connection · First report usually within a day · No card needed for the report