MetadevXInnovate. Code. Create.MetadevXInnovate. Code. Create.

(01) The challenge

Nobody could say what the cloud was costing them

Ultra Tech had grown into forty-one AWS accounts with no shared tagging, three separate logging setups, and an invoice that arrived as a surprise. Incidents were diagnosed by SSH, and cost reviews were arguments about who owned what.

  • · 41 accounts, no consistent tags or ownership
  • · Mean time to detect incidents: 47 minutes
  • · Monthly spend variance of 22% with no explanation

(02) Approach

One pipeline, one board, one owner per resource

We rebuilt the foundation on Terraform so every account inherits the same logging, metrics and tag policy from the start. Grafana became the single pane - infrastructure, application and cost on the same screen - and every spend line resolves to a named team.

  • · Terraform modules with mandatory ownership tags
  • · Centralised logs and metrics with retention tiers
  • · Cost dashboards wired to the same labels as alerts

(03) Outcome

32% off cloud spend in two quarters

Right-sizing and orphaned-resource cleanup came first, then commitment planning against real utilisation. Detection time fell from 47 minutes to under 6, and the monthly cost review became a five-minute read instead of a workshop.

-32%

Cloud spend, two quarters

6 min

Mean time to detect

41

Accounts standardised

(04) What we did

Standards before tooling

We wrote the tagging and ownership policy with finance and engineering in the same room, then made the infrastructure code enforce it. Tooling without that agreement is how the previous three attempts stalled.

  • · Policy workshops with engineering, finance and security
  • · Terraform module library with drift detection
  • · Grafana dashboards for ops and cost side by side
  • · FinOps cadence: weekly anomaly review, monthly commitment check
Next projectRetail E-commerce Rebuild→