LinkedIn Carousel: Tri-Cloud Deployment

Post type: Technical · 10 slides
Blog post: https://www.nuvikatech.com/blog/posts/tri-cloud-deployment

SLIDE 1 — Cover

Headline: We built a platform to help companies control cloud costs. Then deployed it across all three clouds simultaneously.

Sub-line: Yes, we see the irony. Here’s why it’s deliberate — and what it costs.

SLIDE 2

Label: THE ARCHITECTURE

Headline: Dev → GCP. QA → Azure. Prod → AWS. Each environment on a different cloud — on purpose.

Body: Not the result of one design session. Three separate decisions, each made for a specific reason.

SLIDE 3

Label: WHY GCP FOR DEV

Headline: Cloud Run scales to zero. No traffic = no bill.

Body: An equivalent GKE setup would cost ~$800/month in idle node charges. Cloud Run dev costs ~$60/month. For a pre-revenue startup building a cost optimisation platform, this was the obvious call.

SLIDE 4

Label: WHY AZURE FOR QA

Headline: We build Azure scanning. If QA isn’t on Azure, we’re testing against mocks.

Body: When we shipped SLA breach auto-filing for Azure, we ran the first end-to-end test against our own QA subscription. We found one bug that only reproduced against the real Azure Support API.

SLIDE 5

Label: WHY AWS FOR PROD

Headline: Most enterprise customers run significant AWS workloads. Prod should match customer environments.

Body: Network proximity to customer AWS accounts, IAM patterns they recognise, real-world testing of our AWS scanner against our own production account. And we file our own SLA claims there.

SLIDE 6

Label: THE CI/CD COST

Headline: 3 clouds = 3 deployment pipelines, 3 auth systems, 3 monitoring stacks.

Body: WIF for GCP, OIDC for Azure, OIDC for AWS. Every engineer learns all three. Onboarding takes longer. Incident debugging requires knowing which layer failed and how to query that provider’s logs.

SLIDE 7

Label: THE PRODUCT TESTS ITSELF

Headline: Every environment scans the other two. Continuously.

Body: GCP Dev scans Azure QA and AWS Prod. Azure QA scans GCP Dev and AWS Prod. AWS Prod scans GCP Dev and Azure QA. If our AWS scanner has a regression, two environments report failures within the hour.

SLIDE 8

Label: WHAT IT CATCHES

Headline: Cross-cloud blind spots that only appear from a foreign environment’s perspective.

Body: When we built FOCUS 1.2 normalisation, we tested GCP billing data ingested by the Azure scanner. Field mapping issues that looked correct in isolation surfaced immediately when viewed cross-cloud.

SLIDE 9

Label: THE HONEST ADVICE

Headline: Don’t start here. Expand to more clouds when you have a specific reason to.

Body: Dev + QA + Prod across three clouds: ~$180/month plus real operational overhead. We accepted it because the business case is sound. “Avoiding vendor lock-in” alone is not a reason.

SLIDE 10 — CTA

Headline: Want the full story?

Body: The full architecture, the CI/CD patterns, the real cost — and the cross-scanning topology that makes the product continuously test itself.

Link: nuvikatech.com/blog/posts/tri-cloud-deployment