Dev.to Security πŸ” Cybersecurity πŸ‘ 0 πŸ“– 14 min read

How to Standardize Deployment and Governance Across Dozens of Cloud Accounts: A Control-by-Cloud Matrix and What to Do First

Standardize dozens of cloud accounts by putting every account under one organization hierarchy with automated account vending, preventive guardrails, central identity, and central logging, then making every team deploy t

Standardize dozens of cloud accounts by putting every account under one organization hierarchy with automated account vending, preventive guardrails, central identity, and central logging, then making every team deploy through the same pipeline and policy checks. Do it in that order, because guardrails and identity are cheap to set before teams build and expensive to retrofit.

This guide maps each governance control to its AWS, Azure, and Google Cloud equivalent, gives the order to roll them out, and ends with a scorecard for checking whether governance is working.

Disclosure: Control Plane publishes this guide.

TL;DR

Set up the account structure first, then identity, logging, and guardrails, then one deployment path that every team uses.

  • Hierarchy. Put every account, subscription, and project under one organization with a folder or OU structure.
  • Vending. Automate account creation so every new account inherits guardrails from birth.
  • Guardrails. Apply preventive policies that block noncompliant resources before they exist.
  • Identity. Federate all human and workload identity through one provider with short-lived credentials.
  • Logging. Send every account’s audit log to a central, write-protected store.
  • One pipeline. Make every team deploy through the same CI pipeline with policy-as-code checks.
  • The order matters. Hierarchy, identity, and logging first. Detective controls and cost tags last. The table in this guide explains why.
  • Where Control Plane fits. Control Plane standardizes how workloads are deployed, accessed, and audited across AWS, GCP, Azure, and Oracle Cloud Infrastructure (OCI) from one API, one permission model, and one audit trail, complementing the cloud-native governance controls described here.

What is the best way to standardize deployment and governance across dozens of cloud accounts?

The best way is to treat governance as infrastructure: define it in code, enforce it automatically, and apply it uniformly across every cloud provider you use.

Each cloud provider implements governance primitives differently. AWS uses Organizations with OUs and SCPs. Azure uses management groups and Azure Policy. Google Cloud uses a resource hierarchy with folders and Organization Policy constraints. None of these tools interoperate, so teams operating across clouds end up managing three separate governance stacks with three separate mental models.

A control-by-cloud matrix solves this by mapping each governance intent to the specific service or mechanism on each provider. This gives platform teams a single reference for what needs to exist everywhere, regardless of which cloud a workload runs on. It also reveals where a multi-cloud or open-source tool can bridge the gap.

Control AWS Microsoft Azure Google Cloud Multi-cloud or open-source option
Account hierarchy Organizations with OUs Management groups Resource hierarchy with folders No direct equivalent
Landing zone and account vending Control Tower Account Factory, Account Factory for Terraform Azure Landing Zones, subscription vending Fabric FAST, project factory No direct equivalent
Preventive guardrails SCPs and resource control policies Azure Policy deny effects Organization Policy constraints OPA, Sentinel, Checkov
Detective controls AWS Config, Security Hub CSPM Azure Policy compliance, Defender for Cloud Security Command Center Cloud Custodian, Prowler, Steampipe compliance mods
Central identity IAM Identity Center Microsoft Entra ID with PIM Cloud Identity, Workforce Identity Federation External IdP with OIDC federation
Central logging CloudTrail organization trail Azure Monitor diagnostic settings to central Log Analytics workspace Aggregated log sinks SIEM ingestion through Fluentd or the OpenTelemetry Collector
Network baseline Transit Gateway, VPC sharing Hub-and-spoke VNet, Virtual WAN Shared VPC No direct equivalent
Cost allocation Tag policies, Cost Explorer Azure tags, Cost Management Labels, billing export FinOps tooling that ingests FOCUS-format billing exports
Deployment pipeline CI with OIDC federation CI with OIDC federation CI with OIDC federation Terraform/OpenTofu modules plus OPA, Sentinel, or Checkov in CI

Most controls have a direct equivalent on each cloud, but the enforcement mechanism differs. Preventive guardrails are a good example: SCPs set the maximum permissions available to IAM principals in member accounts (they grant nothing and do not apply to the management account), while Azure Policy deny effects evaluate the properties of create and update requests through Azure Resource Manager, and GCP Organization Policy constraints restrict specific resource configurations. All three achieve the same goal through different means. Your governance documentation should describe the intent once and reference the provider-specific implementation for each cloud.

Equivalent cloud services across AWS, Azure, and Google Cloud are compared service by service in a separate guide.

What should you standardize first?

The order in which you roll out governance controls matters as much as which controls you choose. Some controls are prerequisites for others, and some become more expensive to add after teams have already built on top of ungoverned accounts.

Each control in the rollout order is listed with its reason and the cost of adding it later.

Order Control Why it comes at this point Cost of doing it later
1 Organization hierarchy Everything else attaches to this structure. Guardrails, logging, and identity all inherit from the hierarchy. Restructuring OUs or management groups after hundreds of accounts exist requires migration and risks breaking inherited policies.
2 Central identity and SSO Teams need to log in before they can build. Federating identity early prevents local IAM users from proliferating. Create two vaulted break-glass accounts at the same time. Hunting down and rotating local credentials across dozens of accounts is a multi-quarter project.
3 Central logging Audit trails must exist from day one, in dedicated log-archive and security accounts, with security services delegated there and no workloads in the management account. Logs cannot be generated after the fact. Gaps in audit history cannot be backfilled for auditors.
4 Preventive guardrails Block noncompliant resources before they exist. Guardrails are easy to apply to empty accounts and disruptive to apply to full ones. Existing noncompliant resources must be individually remediated or exempted, creating a long tail of tech debt.
5 Account vending Automated vending applies controls 1 through 4 to every new account at creation. Until vending exists, freeze new account creation. Manual account creation leads to inconsistent baselines and shadow accounts.
6 Network baseline Shared networking must be in place before workloads connect to each other or to on premises systems. Retrofitting network architecture around running workloads means renumbering overlapping CIDR ranges, which requires downtime. Allocate address space centrally from the start.
7 Deployment pipeline and policy checks With accounts, identity, and networking in place, teams can start deploying through a governed pipeline. Teams that have already built their own pipelines resist migrating, and ungoverned deployments accumulate.
8 Detective controls and remediation Detective controls catch drift from the baseline established by preventive guardrails. They are most useful once there is a baseline to drift from. Without preventive controls first, detective controls generate overwhelming volumes of findings with no clear remediation path.
9 Cost allocation tags Owner, environment, and cost-center tags belong on the account at vending; resource-level tag enforcement comes last. Retroactively tagging thousands of untagged resources is tedious and error-prone.

Retrofitting existing accounts

Existing estates reverse one step of the rollout order: turn on detective controls first, in report-only mode, to inventory what already exists, because you cannot restrict resources you have not found. Treat the findings as an inventory first and a remediation queue later. Catalog every account, subscription, and project. Classify each by environment, owner, data sensitivity, and compliance requirements. Move accounts into your organization hierarchy in waves, starting with non-production and lower-risk workloads. Apply preventive guardrails in audit or dry-run mode first (Azure Policy audit effects, GCP Organization Policy dry run; for SCPs, check CloudTrail and IAM last-accessed data), review the findings, remediate violations, then switch to enforcement. This phased approach avoids breaking running workloads while closing governance gaps systematically.

How do you standardize deployment across accounts and clouds?

The goal is one pipeline, one set of policy checks, and one way to get credentials, regardless of which cloud a workload targets.

Reusable infrastructure-as-code modules are the foundation. Define modules by intent, such as β€œcreate a production-ready database” or β€œprovision a web-facing load balancer,” and implement each module for the relevant cloud providers using Terraform, OpenTofu, or Pulumi. Teams consume modules from a shared registry and customize them through variables instead of forking the module.

Policy checks belong in CI, between the plan and apply stages. Tools like OPA, HashiCorp Sentinel, and Checkov evaluate planned infrastructure changes against your governance rules and reject noncompliant deployments before they reach any cloud. CI checks can be bypassed, so cloud-side preventive guardrails remain the backstop. Policy as code for multi-cloud IaC governance covers the tooling in detail.

OIDC federation eliminates stored cloud keys from CI pipelines. Most major CI systems, including GitHub Actions, GitLab CI, and Azure Pipelines, issue OIDC tokens, and every major cloud provider accepts them. Your pipeline authenticates to AWS, Azure, or GCP with a short-lived token scoped to the specific account and permissions needed for that deployment. Scope each cloud trust policy to the repository and branch or environment claim, because a wildcard subject lets any repository assume the role.

Per-cloud landing zones like AWS Control Tower and Azure Landing Zones govern a single cloud and do not coordinate with each other. If you run workloads on more than one cloud, you need a layer above the landing zones that enforces common policies and provides a unified view. One identity provider such as Entra ID or Okta can federate into all three clouds.

How do you get a single view across all cloud accounts and regions?

A single pane of glass across cloud providers requires centralizing four categories of data: inventory, logs, cost, and security findings.

Inventory. You need a continuously updated catalog of every account, resource, owner, and configuration state. Each cloud offers native inventory tools: AWS Config records resource configurations, Azure Resource Graph indexes resources across subscriptions, and Google Cloud Asset Inventory tracks resources across projects. For multi-cloud visibility, open-source tools like CloudQuery and Steampipe normalize inventory data from multiple providers into a queryable format.

Logs. Central logging means routing all audit and operational logs to a single store. AWS CloudTrail organization trails, Azure Monitor diagnostic settings, and Google Cloud aggregated log sinks each centralize logs within their respective clouds. For cross-cloud aggregation, organizations typically route logs to a SIEM or a centralized observability platform.

Cost. AWS Cost Explorer, Azure Cost Management, and Google Cloud billing export each provide cost visibility within their cloud. Multi-cloud cost visibility requires aggregating billing data into a FinOps tool that normalizes pricing models, discount structures, and tagging conventions.

Security findings. AWS Security Hub CSPM, Microsoft Defender for Cloud, and Google Cloud Security Command Center each surface security findings natively, and Defender for Cloud can also connect AWS and GCP accounts. Cloud Security Posture Management (CSPM) tools aggregate findings across providers and map them to common frameworks. Categories of CSPM tools include cloud-native options from each provider, standalone multi-cloud CSPM platforms, and open-source scanners that run in CI.

No native tool unifies inventory, logs, cost, and findings across all three clouds. The practical approach is to standardize the data format and destination for each category, then build dashboards and alerts on top of the aggregated data.

How do you know governance is working?

Governance is only effective if you can measure it. Each question has a good answer and a warning sign.

Question Good answer Warning sign
Can anyone create an account outside the vending process? No. All account creation goes through the automated vending pipeline. Teams request accounts through tickets or create them manually.
Do any accounts have local IAM users? No local users except documented break-glass accounts, which are vaulted, MFA-protected, and alert on every use. Local IAM users exist with long-lived access keys.
Is every account’s audit log in a central, write-protected store? Yes, with tamper-resistant retention that meets your compliance requirements. Some accounts have logging disabled or logs stored only locally.
Can a team deploy without the shared pipeline? No. Humans have read-only production access; only the pipeline’s federated role can write. Teams routinely deploy through the console or personal scripts.
What percentage of resources have required tags? Nearly all, enforced by policy at deployment time. Tags are optional, and untagged spend grows each month.
How quickly are policy violations detected and remediated? Within hours, through automated detection and a named owner for each finding. Violations discovered during quarterly audits or incident response.
How many open exceptions exist, and how old are they? A short list, each with a documented justification, an owner, and an expiration date. Permanent exceptions with no review schedule.
Are guardrails applied uniformly across all accounts? Yes. Every account inherits guardrails from the organization hierarchy. Production accounts have guardrails, but dev and sandbox accounts do not.
Is identity federation the only authentication path? Yes, apart from monitored break-glass accounts. No static keys and no service account key files. Some workloads authenticate with static credentials stored in environment variables.
Do policy changes go through version control and review? Yes. All policy changes are pull requests reviewed by the platform team. Policy changes are made directly in the console by individual engineers.

Review this scorecard monthly. Trends matter more than snapshots: a compliance rate that slips a little each month signals a process failure even when the current number looks acceptable.

How Control Plane fits

Control Plane is the unbreakable platform for running production workloads.

Control Plane standardizes how workloads are deployed, accessed, and audited across clouds. It does not replace the cloud providers’ own organization policies, landing zones, or billing for the accounts themselves.

A Control Plane org holds fine-grained permission policies for every resource type, binding users, groups, service accounts, and workload identities to specific permissions. Control Plane’s Universal Cloud Identity (UCI) lets workloads reach AWS, GCP, and Azure services through one identity, with no long-lived cloud keys in your code. One identity model for multiple clouds explains the pattern.

A Control Plane cloud account connects an org to AWS, Azure, or GCP through a provider-side role or service principal with minimal access. A workload deployed to a Control Plane Global Virtual Cloud (GVC), a named set of locations, runs in every location of that GVC from one workload definition. Locations can be Control Plane-managed regions on AWS, GCP, Azure, and Oracle Cloud Infrastructure (OCI), Kubernetes clusters you already run joined through Control Plane’s Bring Your Own Kubernetes (BYOK), or your own servers joined through Control Plane’s Managed Kubernetes (MK8s).

Control Plane provides a tamper-proof audit trail that captures every native Control Plane action. Workloads and third-party systems can write their own entries through an audit context. The CLI, console, Terraform and Pulumi providers, and the Control Plane MCP server all work through the same Control Plane API. Because Control Plane ships its own Terraform and Pulumi providers, it complements the infrastructure-as-code you already run.

Control Plane is PCI DSS Level 1, SOC 2 Type II, HIPAA, and GDPR compliant. Control Plane handles Day 2 operations for you, including autoscaling, built-in observability, and security.

For organizations looking to give developers self-service with guardrails without expanding the platform team, or CTOs looking to reduce the number of cloud tools and vendors, Control Plane runs them from one API and one permission model.

FAQ

Common questions about multi-account governance, landing zones, and guardrails.

What is the best way to standardize deployment and governance across dozens of cloud accounts?

Put every account under one organization hierarchy with automated account vending, preventive guardrails, central identity, and central logging. Then require every team to deploy through the same CI pipeline with policy-as-code checks. The control-by-cloud matrix in this guide maps each governance control to its AWS, Azure, and Google Cloud equivalent so you can enforce the same intent on every provider.

How can I get a single pane of glass view across all our cloud providers and regions?

Centralize four categories of data: inventory, logs, cost, and security findings. Use each cloud’s native aggregation tools to centralize within that cloud, then route the data to a multi-cloud observability platform, SIEM, or FinOps tool that normalizes across providers.

What is a landing zone?

A landing zone is a pre-configured, governed environment in a cloud provider that includes an account hierarchy, identity federation, logging, networking, and guardrails. AWS Control Tower, Azure Landing Zones, and Google Cloud Fabric FAST are examples. Landing zones give every new account a compliant baseline from the moment it is created.

What is account vending?

Account vending is the automated process of creating new cloud accounts, subscriptions, or projects with all required governance controls already applied. AWS Control Tower Account Factory, Azure subscription vending, and Google Cloud project factory are provider-native implementations. The goal is to eliminate manual account setup and ensure every account inherits guardrails, logging, identity federation, and network connectivity from birth.

How do SCPs, Azure Policy, and GCP Organization Policy compare?

All three are preventive guardrails that block noncompliant actions, but they operate differently. AWS Service Control Policies (SCPs) set the maximum API permissions available to principals in member accounts. Azure Policy deny effects evaluate resource properties at deployment time and reject noncompliant configurations. GCP Organization Policy constraints restrict specific resource settings, such as allowed resource locations or disabling service account key creation. The intent is the same: prevent noncompliant resources from being created. The enforcement mechanism and policy language differ.

How do I stop teams creating cloud accounts outside governance?

Remove the ability to create accounts manually. Grant account creation only to the vending pipeline’s identity: IAM in the AWS management account, billing-account roles in Azure, and the Project Creator role in Google Cloud (remove the organization-wide default grants of Project Creator and Billing Account Creator). Find accounts outside the hierarchy through billing and expense data, because detective controls inside the organization cannot see them.

Should each team or environment get its own cloud account?

Usually, split by workload and environment rather than one account per team. Separate accounts provide blast radius isolation, clearer cost attribution, and simpler IAM policies. Production, staging, and development workloads should each run in their own account. Shared services like logging, networking, and identity should have dedicated accounts. Automated vending makes this practical at scale.

How does Control Plane help govern workloads across many cloud accounts?

Control Plane provides a single operational layer for deploying, accessing, and auditing workloads across AWS, GCP, Azure, and Oracle Cloud Infrastructure (OCI). Org-wide permission policies, Control Plane’s Universal Cloud Identity (UCI) for AWS, GCP, and Azure, and a tamper-proof audit trail apply to every workload regardless of where it runs.

What happens if governance controls break a running workload?

Test preventive guardrails before enforcing them: use audit effects in Azure Policy and dry-run organization policies in Google Cloud, and in AWS, where SCPs have no audit mode, attach new SCPs to a test OU first. Review the findings, identify workloads that would be affected, remediate or create documented exceptions with expiration dates, then switch to enforcement. A formal exception process with business justification, compensating controls, and an accountable owner ensures that governance does not become a delivery blocker while still maintaining accountability.

Talk to the Control Plane team or sign up free to deploy and govern workloads across AWS, GCP, Azure, and Oracle Cloud from one API.

Originally published on the Control Plane blog.

πŸ“° Read the original article on Dev.to Security

Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes β€” full credit and traffic to the original publisher.