Workload-Fit Analysis
For every workload: should it live where it is, move, or split? A scorecard across cost, sovereignty, exit risk and operational cost.
Scorecard · Cost model · RiskCloud Platforms · 04 of 04 · Multi-Cloud
Multi-cloud, only when it's the right answer.
The pitch deck says multi-cloud is always smart. The on-call rotation disagrees. We help you decide, honestly, when a second cloud earns its weight — and then we wire the platform that makes the complexity invisible to your engineers.
3Hyperscalers in production
CrossplaneControl plane of choice
vendor-neutralBy design
What we deliver
The strategy first. The wiring second.
Multi-cloud is a strategy, not a sticker. The wiring only works after the decision does.
Unified Control Plane
Crossplane & Terraform as the API for any cloud. Workload teams use one shape; the platform team maps it to the right provider.
Crossplane · Terraform · OPAFederated Identity
One IdP, workload identity federation everywhere, no long-lived keys. Audit trail consolidated regardless of which cloud the workload lives in.
Entra · IAM IC · WIFNetwork & Connectivity
Megaport / Equinix backbones, private interconnects, a service mesh that doesn't know which cloud it's on. Encryption end-to-end.
Megaport · Cilium · IstioCross-Cloud Observability
One pane of glass — metrics, logs and traces normalized via OpenTelemetry, stored in your tool of choice, alerted on the SLO not the platform.
OTel · Datadog · GrafanaExit-Plan Discipline
Quarterly drill of the "what if we had to leave" scenario per workload. Documented. Tested. Filed. The exit you hope to never need.
Exit plan · DR drill · PortabilityThe honest decision tree
Six questions that actually decide.
If you can't answer "yes" with a straight face to at least three, multi-cloud will cost you more than it saves. We'll say it before the contract.
WORKLOAD · WORTH-IT SCORECARD
Is there a regulatory sovereignty requirement?
Yes
→ multi-region per jurisdiction
Does the workload have a single best-of-breed home?
Yes
→ live there · accept lock-in
Is exit cost > 12 months of full migration cost?
Yes
→ multi-cloud for portability
Does your team have 8+ engineers free to operate it?
No
→ stay single-cloud
Is latency to a specific region a hard SLO?
Yes
→ regional cloud-of-choice
Are you doing it because the board asked?
No
→ politely push back
Multi-cloud is a tax. We'll only sign you up if it earns its rate.
Every additional cloud doubles the runbook, halves the team's depth-of-knowledge, and adds an entire IdP, networking, observability and CI surface area. It's a tax worth paying for sovereignty, exit-risk and best-of-breed services — and a waste for almost everything else.
- 1
Discovery workshop
We walk the workloads, the contracts, and the regulatory map. Two days, then a written report.
- 2
Honest scorecard
Every workload gets a yes / no / split recommendation with cost models behind it.
- 3
Reference architecture
If multi-cloud is the call, you get a target architecture with Crossplane, OPA, OTel and IdP federation baked in.
- 4
Implementation sprints
4–6-week sprints, each shipping a working slice. No 200-page transformation programme.
The abstraction stack
The tools that make three clouds feel like one.
Abstraction is the only way multi-cloud works at scale. The right abstraction is invisible to workload teams.
Control Plane
CrossplaneTerraformPulumiBackstage
Policy & Governance
OPA / GatekeeperKyvernoCloud Custodian
Identity
Entra IDAWS IAM ICWorkload Identity FederationSPIFFE
Network
MegaportEquinix FabricCiliumIstio
Observability
OpenTelemetryDatadogGrafana CloudHoneycomb
Cost & FinOps
VantageCloudHealthSpot.ioCloudZero
Data Portability
IcebergDelta UniFormParquetApache Beam
Delivery
Argo CDFluxGitHub ActionsBuildkite
When multi-cloud earned it
Three workloads, three reasons. One control plane.
Real shipped work. The reasons were specific, the wiring was disciplined.
EU citizen data on local cloud, app plane on the global one.
Workload split: storage and analytics on a sovereign provider, application plane on AWS. Federated identity, OTel telemetry, one runbook.
2Clouds
0Cross-border bytes
CrossplaneOPAWIF
Two clouds, one company, six months in.
Acquired company on Azure, parent on AWS. Identity federation in week two, network peering in week six, single Backstage portal by month four.
6 moTo single portal
0Workload re-platforms
BackstageEntraMegaport
App on AWS. Model on Vertex. Warehouse on Snowflake.
Each component in its sharpest home, glued via private interconnect and a shared semantic layer. One SLO surface, three sets of receipts.
3Specialty platforms
1SLO dashboard
VertexSnowflakeOTel
The honest multi-cloud conversation.
30 minutes. Bring the workloads you're wondering about — we'll tell you whether multi-cloud actually earns its tax in your case.