Skip to main content

Glossary

The terms below are used the same way on every page of this site and in the repositories of every Baseline edition. Each definition is one or two sentences with a link to the page that explains the term in full. Shared terms come first, then the terms that belong to the AWS, Azure and GCP Enterprise Baselines. The key concepts page covers the ten ideas to learn first, the AWS services used page indexes the AWS service explainers, the Azure services used page indexes the Azure service explainers, and the GCP services used page indexes the GCP service explainers.

How the core terms relate​

Shared terms​

Account, subscription or project. The isolation unit of each edition: the hard boundary for access, billing and blast radius. The AWS Enterprise Baseline runs on 15 accounts, the Azure Enterprise Baseline on 15 subscriptions and the GCP Enterprise Baseline on 15 projects. See landing zones.

Baseline unit. The settings every isolation unit receives from one unit: account-baseline on AWS (default EBS encryption, S3 public access block, AWS Config, the deployer and reader roles), subscription-baseline on Azure (resource providers, Defender plans, Activity Log export, Network Watcher, a locked resource group) and project-baseline on GCP (APIs, a de-privileged default compute account, the deployer, log retention, OS Login). See subscriptions for the Azure one and Resource Manager for the GCP one.

Blueprint. A production-ready reference architecture for a real workload, delivered as its own repositories and deployed into the stage accounts, subscriptions or projects. Each edition has three: Web App, Data and ETL, and Secrets. See the blueprints.

core and plat. The two halves of every edition's tree. core holds the shared, single-purpose units (security, audit, identity, network, dns, artifacts, auto, corp, public, backup) and plat one unit per stage. They are organizational units on AWS, the management groups mg-acme-core and mg-acme-plat on Azure, and folders on GCP, where the backup project sits in a third top-level folder, backup, of its own. See landing zones.

Deployer identity. The identity a pipeline uses to apply changes. A blueprint repository has one per stage, which accepts only jobs in that stage's GitHub Environment; the Baseline has one, which accepts only the prod Environment. On AWS it is an IAM role such as acme-aws-blueprint-webapp-infra-dev-deployer, chaining into acme-terraform-deployer in the target account; on Azure a federated managed identity such as id-acme-azure-blueprint-webapp-infra-dev-deployer; on GCP a CI service account such as sa-acme-webapp-infra-dev-ci that, for an infrastructure repository, impersonates the target project's sa-acme-terraform-deployer. See least privilege.

Drift. A difference between what the code declares and what exists in the cloud. A scheduled Baseline workflow plans the core-critical and active stage units Tuesday to Friday and every unit on Monday (on Azure, every unit on Monday by default), and reports any difference. See drift.

Entitlement. The products an organization holds, which decide the documentation sets its members can open. See how to use these docs.

Environment. The region slug a folder deploys to: usw2 for us-west-2, eus2 for eastus2 or usw1 for us-west1, or glob for the global folders that hold organization-wide singletons. See environments and promotion.

Generated tree. The git-ignored .terragrunt-stack/ folder that Terragrunt produces from a stack file. Nothing in it is hand-written. See Terragrunt units and stacks.

GitOps. Keeping the desired state of a system in Git and letting a controller reconcile the running system to it. In the Web App Blueprints the controller is ArgoCD. See GitOps.

Guard script. A small Python check in a repository's scripts/ folder that enforces a layout or configuration rule before merge, such as no hard-coded CIDRs or mocked outputs on every dependency. See policy as code.

Home region. The region that holds the global units and the organization-wide resources, set once in common.hcl before the first deployment. The samples are us-west-2, eastus2 and us-west1. See environments and promotion.

Hub. The shared network every stage reaches the internet and the shared services through: the hub VPC in core-network on AWS, the hub VNet in acme-core-network on Azure, and the hub VPC of the Shared VPC host acme-core-network on GCP. See hub-and-spoke networking.

Immutable artifact. A build output, such as a container image tag, that is never overwritten. The same artifact moves from dev to prod unchanged. See immutable artifacts.

Isolation domain. A named group of stages that may reach each other but no other group, such as prod and nonprod. It is a Transit Gateway route table on AWS, a rule collection of the hub Azure Firewall policy on Azure, and a separate VPC peered only with the hub on GCP. See hub-and-spoke networking.

Landing zone. The multi-account, multi-subscription or multi-project foundation applications run on: isolation units, network, security services, identity, and the delivery machinery that manages them. Each Baseline edition is a landing zone. See landing zones.

Landing-zone contract. The values a Baseline publishes for each stage so the blueprints never look up or hard-code them: Systems Manager parameters under /acme/<environment>/<stage>/ on AWS, 30 secrets in the contract vault kv-acme-eus2-<stage>-plat on Azure, and 24 Parameter Manager parameters in each stage project on GCP. See Landing-zone contract for the concept, Key Vault for the Azure contract vault and Parameter Manager for the GCP parameters.

Least privilege. Granting each person and pipeline only the access its job needs, and narrowing it toward production. See least privilege.

Mock outputs. Placeholder values a unit declares for each dependency, so a new environment can be planned end to end before its dependencies exist. See Terragrunt.

Module. An OpenTofu module: a folder of resources with inputs and outputs. A unit instantiates one module. See OpenTofu.

Namespace. The short organization abbreviation that starts every resource name, such as acme in acme-usw2-prd-vpc, acme-eus2-prd or acme-plat-dev. See how to use these docs.

Placeholder value. A sample such as acme, company.com, 111111111111 or eastus2 that stands for a value your organization sets. Signed-in readers see their own values. See how to use these docs.

Promotion. Moving a change from dev through staging to prod without rebuilding it. See environments and promotion.

Remediation. An automated response to a finding or a misconfiguration. The AWS edition blocks public access on an S3 bucket that becomes public outside the public account; the GCP edition's optional Security Command Center notifier removes public bindings from a bucket outside acme-core-public; the Azure edition prevents the change up front with Azure Policy and repairs missing tags with a Modify effect. See Azure Policy.

Spoke. A stage's own network, attached to the hub: a spoke VPC on AWS, a spoke VNet peered with the hub VNet on Azure, and on GCP the stage's subnets in its isolation-domain VPC, used by the stage project as a service project. See hub-and-spoke networking.

Stack file. A terragrunt.stack.hcl file in an environment folder that lists the units the folder deploys and their values. See Terragrunt units and stacks.

Stage. The role of a stage account, subscription or project and its short code: sbx, dev, stg and prd. See environments and promotion.

State file. The record OpenTofu keeps of what it created for one unit. State lives in an S3 bucket in the management account on AWS, in the tfstate container of the storage account stacmeeus2roottfstate in acme-core-root on Azure, and in the bucket acme-usw1-root-tfstate of the seed project on GCP. See OpenTofu and Terraform.

Switch file. A small file that holds the feature switches of one area, each with its price, so a decision is one reviewed line: network.hcl per region and security.hcl in every edition, plus identity.hcl and plat.hcl on GCP. See Microsoft Defender for Cloud for one in use.

Template. A reusable stack under stacks/ that a stack file includes for every stage: plat-account on AWS, plat-subscription on Azure, plat-project on GCP. See Terragrunt units and stacks.

Unit. One deployable module instance with one state file, defined once in units/<unit>/terragrunt.hcl. See Terragrunt units and stacks.

Values. The per-instance inputs a stack file passes to a unit, so one unit definition serves every account, subscription or project that lists it. See Terragrunt units and stacks.

AWS terms​

Baseline node group. The EKS managed node group that carries Kubernetes and the system-critical workloads, tainted so that nothing else lands on it and sized by Cluster Autoscaler. Application pods run on Karpenter capacity instead. See EKS.

Conformance pack. A bundle of AWS Config rules evaluated together against a standard such as CIS, NIST or HIPAA. The AWS Baseline ships the pack files and deploys the ones switched on. See AWS Config.

Delegated administrator. A member account that AWS Organizations allows to administer a service for the whole organization, so the management account does not have to. The security services are delegated to the security account. See GuardDuty.

Management account. The account that owns the AWS Organization and billing. The AWS Baseline calls it core-root and keeps workloads out of it. See AWS Organizations.

Permission set. An IAM Identity Center template that becomes an IAM role in each account it is assigned to. People sign in through permission sets, never as IAM users. See IAM Identity Center.

Service control policy (SCP). An AWS Organizations policy that sets the maximum permissions for every identity in the accounts it is attached to. The AWS Baseline attaches three. See service control policies.

SSM parameter. A value in Systems Manager Parameter Store. The landing zone publishes VPC, subnet, Transit Gateway attachment and account values under /acme/<environment>/<stage>/ and the blueprints read them. See Systems Manager Parameter Store.

Azure terms​

Action group. The Azure Monitor object that sends an alert to receivers. Every security alert of the Azure Enterprise Baseline routes to one, ag-acme-security-alerts, which emails the security mailbox. See Azure Monitor.

Activity Log. The control-plane log of a subscription: every create, update, delete and role change. Each subscription exports it to the audit workspace and the immutable audit storage in acme-core-audit. See subscriptions.

Azure Policy initiative. A set of policy definitions assigned together. The Azure Baseline assigns its security guardrails, region restriction, audit protection, tag policy and compliance initiatives at mg-acme. See Azure Policy.

Azure RBAC. Azure role-based access control: a role granted to a principal at a scope. People receive roles only through Entra groups, from one access matrix. See Azure RBAC.

Break-glass group. ACME_BreakGlass, the group of emergency accounts excluded from every Conditional Access policy; every sign-in by one raises an alert. See Conditional Access.

Conditional Access. The Microsoft Entra ID policy engine that requires MFA or blocks a sign-in. The Azure Baseline defines five policies, report-only until enforced. See Conditional Access.

Contract vault. The per-stage Key Vault kv-acme-eus2-<stage>-plat that holds the Azure landing-zone contract, 30 secrets named <area>--<name>. See Key Vault.

Customer-managed key (CMK). A Key Vault key you control that encrypts a service's data. The audit storage is encrypted with audit-storage, an RSA 3072 key rotated yearly. See Key Vault.

Defender plan. A Microsoft Defender for Cloud protection plan for one resource type, set to Standard or Free in every subscription from security.hcl. See Microsoft Defender for Cloud.

DenyAction. An Azure Policy effect that blocks one action, such as delete, on matching resources. The audit-protection initiative uses it for the audit workspace, the audit storage and diagnostic settings. See Azure Policy.

Deployment Safeguards. The AKS feature that checks workloads against Azure Policy best practices. The Web App Blueprint's cluster runs it in Warn mode. See Azure Kubernetes Service.

DoNotEnforce. The Azure Policy assignment mode that evaluates and reports but applies no effect. The built-in compliance frameworks are assigned this way. See Azure Policy.

Egress mode. The hub_egress switch in network.hcl: azure_firewall (the default) routes every spoke's internet traffic through the hub firewall, spoke_nat gives each spoke subnet a NAT gateway. It is chosen before the first blueprint deployment. See NAT Gateway and Azure Firewall.

Federated identity credential. The trust on a managed identity for one external token issuer and exact subject, such as a GitHub workflow or a Kubernetes service account. See GitHub OIDC.

Firewall policy. The rule set of the hub Azure Firewall: platform allow rules, VPN access, the isolation domains and the internet egress rules, in one policy. See Azure Firewall.

Log Analytics workspace. The Azure Monitor store for logs, queried with KQL (Kusto Query Language). The Azure Baseline has an audit workspace, a security workspace and one application workspace per stage. See Log Analytics.

Managed identity. A Microsoft Entra principal whose credentials Azure manages. Every pipeline, controller and workload of the Azure platform has its own. See managed identities.

Management group. A container above subscriptions whose policies and roles every subscription beneath it inherits: mg-acme, with mg-acme-core and mg-acme-plat below. See management groups.

Microsoft Entra ID. The identity provider of the Azure tenant: users, groups, service principals and managed identities. See Microsoft Entra ID.

Node Auto Provisioning. The AKS mode that creates nodes of the right size for pending pods. The Web App Blueprint's cluster uses it for application capacity. See Node Auto Provisioning.

PIM (Privileged Identity Management). The Entra feature that makes a role or a group membership eligible and time-boxed instead of standing. The Azure Baseline uses PIM for Groups for Owner and Security Admin behind enable_pim: each role sits on an elevation group whose membership the access group activates. See Privileged Identity Management.

Private endpoint. A network interface in a subnet that gives a PaaS (platform as a service) resource, such as a vault or a database, a private address. See Private Link.

Private Link zone. A private DNS zone such as privatelink.vaultcore.azure.net that resolves a resource's name to its private endpoint. The Azure Baseline holds ten of them in acme-core-network. See Azure Private DNS.

Resource group. The container of a unit's Azure resources; each unit owns its own. See resource groups and locks.

Resource lock. A management lock; CanNotDelete refuses every delete until it is removed. See resource groups and locks.

Service tag. A Microsoft-maintained name for the address ranges of an Azure service, such as AzureDatabricks, used in firewall and network security rules. See Azure Firewall.

Subscription. The Azure isolation unit for billing, quotas and access. See subscriptions.

Unity Catalog. The Azure Databricks governance layer for catalogs, schemas and tables. The Azure Baseline holds its metastore root storage in acme-core-artifacts, and the Data and ETL Blueprint grants on it per stage. See Unity Catalog.

VNet peering. A private connection between two virtual networks. Every stage spoke and the runner network peer with the hub. See virtual network peering.

Workload ID (Microsoft Entra Workload ID). Federation between a Kubernetes service account and a managed identity, so a pod gets Entra tokens without a secret. See managed identities.

GCP terms​

Aggregated sink. A log sink at the organization that routes logs of every project beneath it. The GCP Baseline has two: to the central log bucket and to the archive bucket in acme-core-audit. See Cloud Logging.

Artifact Analysis. The Google Cloud service that scans container images in Artifact Registry for known vulnerabilities, on push and continuously. It is on in the GCP Baseline. See Artifact Analysis.

Cloud Armor rule template. The GCP Baseline's policies/cloud-armor-template.json, published to each stage as the parameter waf--policy-template, from which the Web App Blueprint builds its Cloud Armor security policies. See Google Cloud Armor.

Cloud NAT. The managed outbound NAT of a VPC. Each GCP VPC has its own Cloud Router and Cloud NAT per region, and no VM has an external IP. See Cloud NAT.

Folder. A node of the Google Cloud resource hierarchy between the organization and its projects. The GCP Baseline has two, core and plat. See Resource Manager.

Hierarchical firewall policy. A firewall policy attached to the organization or a folder that applies to every VPC beneath it, before the VPCs' own policies. See Cloud NGFW firewall policies.

IAM deny policy. An IAM policy that denies permissions regardless of any allow, Owner included. The GCP Baseline uses them to protect the audit sinks, log buckets and archive from deletion, to keep the privileged service accounts to the platform pipelines and leads, and to keep auditors away from data. See IAM deny policies.

Identity-Aware Proxy (IAP). The Google Cloud service that tunnels TCP to private instances after an IAM check. GCP engineers reach private resources through IAP and OS Login instead of a VPN. See Identity-Aware Proxy.

Impersonation. One principal acting as a service account through a short-lived token, without a key. The CI service accounts of the GCP infrastructure repositories impersonate the deployer of the project they change; the code repositories' accounts never do. See service accounts.

Log bucket. A Cloud Logging store with its own retention. The GCP Baseline's central log bucket in acme-core-audit keeps 365 days and is linked to a BigQuery dataset. See Cloud Logging.

Organization policy (constraint). A restriction on what any principal can configure, inherited by every project. The GCP Baseline sets 19 at the organization node. See Organization Policy Service.

Organization-scoped unit. One of the 15 GCP Baseline units, marked by an org-scoped.hcl file, that act on the organization rather than one project and run as the baseline CI service account without impersonation. See service accounts.

OS Login. The Compute Engine feature that maps SSH access to a user's Google identity and IAM roles, required by organization policy on GCP. See Identity-Aware Proxy.

Parameter Manager parameter. A versioned, non-secret configuration value. The GCP landing-zone contract is 24 parameters named <service>--<key> in each stage project. See Parameter Manager.

Privileged Access Manager entitlement. A grant of a role that a person requests for a limited time, at most 12 hours, with approval where configured. The GCP Baseline uses entitlements for Owner on the folders and for organization administration. See Privileged Access Manager.

Private Service Access. A peered address range through which Google-managed services get private addresses in a VPC. Each GCP VPC has one. See Private Service Access.

Project. The Google Cloud isolation unit for resources, billing and IAM. The GCP Baseline has 15, named acme-<folder>-<name>. See Resource Manager.

Secure tag. A tag bound to a resource that policies can match. The GCP Baseline binds purpose and stage to projects; the IAM deny policy matches purpose=audit. See Resource Manager.

Security Command Center. Google Cloud's organization-wide findings service. The GCP Baseline sends its HIGH and CRITICAL findings to a Pub/Sub topic. See Security Command Center.

Service project. A project attached to a Shared VPC host that uses the host's subnets. Each GCP stage project is one, limited to its own subnets. See Shared VPC.

Shared VPC (host project). The arrangement in which one host project owns the VPCs and shares their subnets with service projects. acme-core-network is the only host. See Shared VPC.

VPC Service Controls (perimeter, dry run). A perimeter that restricts Google API access across its boundary. The GCP Baseline can define one around the stage projects, always written as a dry-run spec and enforced only by a separate switch. See VPC Service Controls.

Workload Identity Federation (pool, provider). Trust between Google Cloud and an external token issuer. A pool with a GitHub provider lets each repository's workflows act as its CI service account without a key. See Workload Identity Federation.

Workload Identity Federation for GKE. The same federation for Kubernetes service accounts, so GKE pods call Google APIs as IAM principals without a key. See Workload Identity Federation and Google Kubernetes Engine.