Skip to content

Multi-Cloud Governance

A cross-cutting reference for architecture and governance patterns that apply when an enterprise runs workloads across two or more public cloud providers -- typically AWS, GCP, Alibaba Cloud, and/or Tencent Cloud, with Azure also common in Western markets.

This topic covers nine governance domains that together form a vendor-neutral operating model for multi-cloud environments.

Governance Domains

1. Multi-Cloud Networking

Three canonical topologies -- hub-spoke, full-mesh, and transit -- each with distinct trade-offs in latency, blast radius, and operational complexity. Key interconnect services include AWS Transit Gateway + Direct Connect, GCP Cloud Interconnect, Alibaba Cloud Express Connect, and Tencent Cloud Direct Connect. Third-party fabric providers (Equinix Fabric, Megaport, PacketFabric) offer software-defined cross-cloud circuits. See infrastructure/multi-cloud-governance/architecture#networking-patterns.

2. Identity Federation (IAM)

A single IdP (Okta, Entra ID, Keycloak) federates to every cloud via SAML 2.0 for human SSO and OIDC for workload identity. Each cloud maps federated assertions to local roles: AWS IAM Identity Center, GCP Workforce/Workload Identity Federation, Alibaba RAM, Tencent CAM. SPIFFE/SPIRE provides a workload-identity overlay independent of any cloud. See infrastructure/multi-cloud-governance/security#identity-federation.

3. DNS and Traffic Management

Global DNS services -- AWS Route 53, GCP Cloud DNS, Alibaba Cloud DNS (Alidns), Tencent DNSPod -- serve as the first hop for multi-region, multi-cloud traffic steering. Each supports latency-based routing, geolocation, weighted round-robin, and health-check failover. GSLB (Global Server Load Balancing) layers like NS1, Cloudflare, or the providers' own traffic-management features unify policy across clouds. See infrastructure/multi-cloud-governance/architecture#dns-and-traffic-management.

4. Observability

OpenTelemetry is the lingua franca: deploy OTel Collectors per-cloud, centralize into a vendor-neutral backend (Grafana LGTM stack, Datadog, Dynatrace). The three signals -- traces, metrics, logs -- plus the emerging fourth signal (continuous profiling) are all carried over OTLP. Cloud-native services (CloudWatch, Cloud Monitoring, SLS, Tencent CLS) feed into the same collector pipeline via their respective OTel exporters. See infrastructure/multi-cloud-governance/operations#observability.

5. Cost Management (FinOps)

FinOps Foundation's Crawl-Walk-Run maturity model governs spend visibility, allocation, and optimization across clouds. Centralized cost aggregation via Apptio Cloudability, Flexera One, or CloudHealth; Kubernetes-granular allocation via Kubecost/OpenCost; AI-driven rightsizing via CAST AI or ProsperOps. Commitment-based savings (RIs, Savings Plans, Reserved Instances) are layered per-provider. See infrastructure/multi-cloud-governance/operations#finops-and-cost-management.

6. Security Baseline

CIS Benchmarks exist for every major cloud (AWS Foundations, Azure Foundations, GCP Foundations). Automated scanning via Prowler (AWS), Azure Security Center, GCP Security Command Center, or cross-cloud tools like Prisma Cloud, Wiz, and Checkov. Policy-as-Code enforcement via OPA/Gatekeeper, HashiCorp Sentinel, or cloud-native org policies (AWS SCPs, Azure Policy, GCP Organization Policy). See infrastructure/multi-cloud-governance/security#security-baseline-and-compliance.

7. Infrastructure-as-Code

Tool Paradigm Multi-Cloud State License
Terraform / OpenTofu Declarative HCL Broadest provider coverage Remote backend (S3, GCS, OSS, COS) BSL / MPL 2.0
Pulumi Imperative code (TS, Python, Go, C#, Java) Same-day new-service support Pulumi Cloud or self-managed Apache 2.0
Crossplane K8s-native CRDs Growing (AWS, GCP, Azure, Alibaba) Kubernetes etcd Apache 2.0

For detailed configuration examples, see infrastructure/multi-cloud-governance/architecture#infrastructure-as-code.

8. CI/CD and GitOps

GitOps (ArgoCD, Flux) provides a pull-based reconciliation loop for Kubernetes workloads across clusters in any cloud. CI pipelines (GitHub Actions, GitLab CI, Jenkins) produce artifacts; GitOps controllers promote them through environments. Multi-cluster patterns use ApplicationSets (ArgoCD) or Kustomize overlays (Flux) with per-cloud value files. Secret management via Sealed Secrets, SOPS, or External Secrets Operator. See infrastructure/multi-cloud-governance/architecture#cicd-and-gitops.

9. Vendor-Neutral Reference Architecture

Analyst frameworks (Gartner CIPS Magic Quadrant, Forrester Wave for Multi-Cloud Management) recommend composable, cloud-native stacks: Kubernetes + Istio/Cilium + OpenTelemetry + OTel Collector + Terraform/Crossplane + GitOps + OPA. The reference architecture is a layered diagram in infrastructure/multi-cloud-governance/architecture#reference-architecture.

Sources

Questions

  • What is the realistic blast-radius trade-off between hub-spoke and full-mesh topologies for an APAC-centric enterprise running AWS + Alibaba Cloud?
  • How do Alibaba RAM role-session-name constraints compare with AWS IAM role trust-policy patterns when federating from a single IdP?
  • Can Crossplane compositions for Alibaba Cloud match the maturity of AWS/GCP compositions for production use, or is Terraform the safer choice for China-cloud resources?
  • What is the current state of OpenTelemetry log signal stability across Alibaba Cloud SLS and Tencent CLS OTLP ingestion?
  • How should an enterprise normalize FinOps unit economics (cost-per-transaction) when the same transaction spans two clouds?