Appendix G

Terminology and Acronyms

Scope. The terms and acronyms this book uses, grouped by domain, each with the section that defines it. A reminder of what a word means, not a place to learn the concept behind it. Prerequisites. None. Verified against. Google Cloud console and API surface as of 2026-09, Cloud SDK 583.0.0; every section reference re-resolved against its owning chapter; see sources at end.

The chapters define each term where it is first needed, which is the right place to learn it and the wrong place to look it up six months later. This appendix is the lookup: one line each, sorted by domain rather than by the order the book happens to introduce them.

The § is the authority. Where a one-line definition and a chapter disagree, the chapter is right and the line here is too short. Acronyms are given with the expansion first, since that is what a reader who does not recognize one actually needs.

G.1 GCP Terminology §

Platform vocabulary: the hierarchy, the resource containers, and the words Google uses for its own building blocks.

TermDefinition§
CELCommon Expression Language, the expression syntax used by IAM conditions, custom organization policy constraints, and federation attribute mappings.§2.30
Control planeThe API surface that creates, modifies, or describes configuration; audited in Admin Activity logs.§1.8
Data planeTraffic that reads or writes the content a resource holds; audited only if Data Access logs are enabled.§1.8
FolderAn optional hierarchy node between organization and project that carries IAM, organization policy, tags, and hierarchical firewall policies. Nests ≤10 deep, ≤300 children per parent.§1.7, §2.8
LabelNon-inherited key/value metadata used for cost attribution, inventory, and filtering; unusable in IAM conditions or organization policy. Max 64 per resource, ≤63 chars.§2.12
Launch stage (GA / Preview / Beta / Alpha)The support label on an API surface or role; Preview and earlier must never be a production dependency.§3.12
Managed instance group (MIG)A set of instances created from one instance template and kept conformant to it; regional by default, stateless unless explicitly configured otherwise.§8.11
OrganizationThe hierarchy root, one per Cloud Identity or Workspace account, provisioned automatically on domain association.§1.7, §2.7
Organization policyA constraint evaluated by the Google Cloud API itself, so it holds regardless of how a change is submitted.§2.30
ProjectThe container owning billing, quota, API enablement, and the default IAM boundary; the blast-radius unit.§1.5, §2.9
Project ID6–30 char immutable globally unique identifier; never reusable after deletion.§2.9
Project numberGoogle-assigned numeric identifier used by service agents and tag bindings.§1.5
Resource ManagerThe API owning organizations, folders, projects, tag keys, tag values, and tag bindings.§2.6
TagInheritable key/value metadata implemented as discrete TagKey, TagValue, and TagBinding resources; usable in IAM conditions and organization policy.§2.13

G.2 IAM Terminology §

Identity, roles, and the federation machinery. The distinctions that cause the most trouble in practice are between a role and a permission, and between impersonating an account and being attached to one.

TermDefinition§
ADCApplication Default Credentials — the client-library search order that finds an attached identity or a federation credential file without a key.§1.10
Attached service accountA service account bound to a running resource so its code obtains short-lived tokens from the metadata server instead of using a downloaded key.§1.10
Attribute conditionCEL on a federation provider that must be true for an assertion to be accepted; where the security decision belongs.§4.1
Attribute mappingCEL turning IdP claims into google.subject, google.groups, and attribute.NAME.§4.1
Basic rolesSix coarse project-wide roles: legacy Owner, Editor, Viewer, plus Admin, Writer, Reader, which are Preview. Never granted in this book.§3.10
Custom roleA role you define; creatable at organization or project only, capped at 300 per org, and does not gain permissions as services add methods.§3.12
Identity-Aware Proxy (IAP)Central authorization layer in front of web resources and administrative TCP tunnels. Application-delivery use is §7.17.§4.16
ImpersonationObtaining a short-lived credential for a service account; logs both the human and the account.§3.24
OIDCOpenID Connect — the token format behind most workload federation providers.§4.5
OS LoginGoverning Linux host access through IAM roles (roles/compute.osLogin, roles/compute.osAdminLogin) rather than metadata SSH keys; §8.19 owns the mechanics.§4.18, §8.19
PrincipalAny identity IAM can name in a policy; the identifier prefix determines the identity type and its governance.§3.2
Privileged Access Manager (PAM)Google Cloud's just-in-time privilege service; GA for core grants, refusing the legacy Owner/Editor/Viewer roles.§3.27
SAMLSecurity Assertion Markup Language — the assertion format used by Workforce Identity Federation and legacy SSO.§4.6
Security Token Service (STS)The sts.googleapis.com endpoint that exchanges a verified external assertion for a short-lived Google credential, per RFC 8693.§4.1
Service account keyA downloadable, non-expiring bearer credential; an anti-pattern throughout this book.§3.25
Service agentA Google-created, Google-managed service account that lets a service act on your behalf.§1.10
WIFThe book abbreviation for Workload Identity Federation.§4.3
Workload Identity FederationExchange of an external workload's native token for a Google credential; the mechanism that removes service account keys.§4.3

G.3 Networking Terminology §

Addressing, routing, and the data path. Terms whose subject is an authorization boundary rather than a path — service perimeters above all — are in G.7 instead, because grouping them with subnets teaches the wrong model.

TermDefinition§
Alias IP / secondary rangeAn additional subnet range used by GKE pods and services so their addresses are routable and visible to firewall policy.§5.4
ASNAutonomous system number, chosen per Cloud Router and per on-premises peer.§5.21
Backend serviceThe load balancer object holding Cloud Armor policy, IAP, CDN, logging, timeouts, and the backend list; where most delivery security configuration lives.§7.6
BGPBorder Gateway Protocol, the dynamic routing protocol Cloud Router speaks over VPN and Interconnect.§6.3
Cloud NATDistributed managed outbound NAT configured on a Cloud Router; not an egress control.§5.20
Host projectThe Shared VPC project owning the network, subnets, routes, firewall policy, Cloud Router, NAT, and DNS.§5.28
Implied rulesEvery VPC's undeletable defaults: deny ingress to VM instances, allow all egress.§5.13
NEGNetwork endpoint group — the backend type addressing endpoints directly rather than instance groups.§7.6
NGFWCloud Next Generation Firewall — the current name for firewall policies and their L7 inspection surface.§5.31
Private Google AccessPer-subnet setting letting VMs without external IPs reach Google APIs over internal paths.§5.17
Private Service AccessVPC peering to a Google service producer network from a range you allocate; not transitive, IPv4 only.§5.18
Private Service ConnectConsumer-side endpoints or backends reaching a producer service or Google API bundle, with independent address spaces on both sides.§5.19
Proxy-only subnetA REGIONAL_MANAGED_PROXY or GLOBAL_MANAGED_PROXY subnet supplying addresses to Envoy proxies; one active and one backup per region per network, minimum /26.§5.4, §7.4
restricted.googleapis.comPrivate VIP range 199.36.153.4/30 reaching only VPC Service Controls–supported APIs; the correct choice inside a perimeter.§1.9
Service projectA project attached to a Shared VPC host, using its subnets while keeping its own IAM, billing, and quota.§5.29
Shared VPCThe arrangement separating network ownership (host project) from workload ownership (service projects).§5.27

G.4 Kubernetes Terminology §

Kubernetes and GKE vocabulary, limited to what this book uses. Kubernetes' own general concepts are assumed.

TermDefinition§
Container-native load balancingGKE load balancing that targets pod IPs directly through zonal Network Endpoint Groups instead of node IPs and a NodePort.§9.23
Container-Optimized OSGKE's default minimal, auto-patching, read-only-root node image.§9.31
Dataplane V2GKE's eBPF-based dataplane built on Cilium, replacing iptables and making Kubernetes NetworkPolicy always-on.§9.16
Direct principal bindingGranting an IAM role straight to a federated identity (e.g. a Kubernetes ServiceAccount) with no Google service account in between.§4.12
DNS-based control plane endpointThe current recommended GKE control-plane access model: a stable FQDN integrated with IAM and VPC Service Controls, enabled with --enable-dns-access.§9.4
GKE EnterpriseThe paid GKE edition bundling fleet-wide Config Sync, Policy Controller, and related features beyond GKE Standard.§9.38, §9.39
GKE metadata serverPer-node component intercepting metadata.google.internal and exchanging Kubernetes ServiceAccount tokens for federated tokens; 500 concurrent connections per node.§4.12
GKE SandboxA gVisor-based per-node or per-pod isolation layer re-implementing the kernel syscall surface in userspace.§9.2
Pod Security AdmissionThe built-in admission controller enforcing the Pod Security Standards (privileged, baseline, restricted), replacing PodSecurityPolicy (removed in Kubernetes 1.25).§9.14
RBACKubernetes role-based access control, evaluated inside the cluster and separate from IAM.§9.12
VPC-native clusterA GKE cluster using alias IP secondary ranges for pods and Services; the only networking mode for new clusters.§9.7

G.5 SecDevOps Terminology §

Practice vocabulary: reliability, delivery, and the measurements that make both arguable. These are the terms an engineer is expected to already know and most often knows only approximately.

TermDefinition§
AttestationA signed statement that an image passed a check, created against a Binary Authorization attestor and bound to the image's digest; the attesting identity must not be the build identity.§25.11
Build configThe cloudbuild.yaml document defining steps, options, substitutions, artifacts, and the service account; it must live in the protected repository, not inline in a trigger.§23.1
Build stepA container image run in order against a shared /workspace; its image is an input to the build and must be pinned by digest.§23.3
CVECommon Vulnerabilities and Exposures identifier, the unit of a vulnerability finding.§24.6
Delivery pipelineThe ordered list of Cloud Deploy stages, each naming a target and optionally a strategy and Skaffold profiles.§25.1
DSSEThe Dead Simple Signing Envelope wrapping a provenance payload — payloadType, payload, signatures.§37.5
Immutable tagsA Docker-format repository setting preventing tags from being deleted or moved to a different digest and preventing tagged images from being deleted.§24.5
OccurrenceA metadata record attached to an artifact — build, discovery, vulnerability, or SBOM reference — surfaced by --show-occurrences.§24.2
Private poolA set of Cloud Build workers in a network peered to yours, optionally with --no-public-egress, which is the only constraint on exfiltration from a build.§23.5
Recovery time objective (RTO)The maximum tolerable time to restore service after failure; a target you set and measure, not a figure Google publishes.§E.9
Remote repositoryAn Artifact Registry mode that fetches from a public upstream and caches it, turning public dependency risk into a store you control.§24.1
SBOMSoftware bill of materials, surfaced as an Artifact Analysis occurrence on an image.§24.2
SLSAThe framework of levels describing a build pipeline's tamper resistance; Cloud Build provenance supports level 3 assurance for the build platform, not for your source practices.§23.14
SLSA trackA dimension of the v1.2 specification; the Build track has levels 1–3 and the published Source track has four, its top level requiring two-person review.§37.10

G.6 Infrastructure-as-Code Terminology §

Terraform, OpenTofu, and Ansible vocabulary, including the few terms where the two Terraform implementations genuinely differ.

TermDefinition§
auth_kindThe required google.cloud credential selector, taking application, machineaccount, serviceaccount, or accesstoken; only serviceaccount consumes a key file.§28.4
Backend blockThe root-module declaration of where state lives; a child module may not have one, and under OpenTofu alone it may reference variables (§27.8).§26.14
Data sourceA read-only reference to infrastructure another configuration owns; some return secrets into state, and provider 8 supplies ephemeral equivalents for the two worst.§26.11
DriftThe difference between refreshed state and what state records; detected with terraform plan -refresh-only -detailed-exitcode, where exit code 2 means a diff.§26.22
Ephemeral resourceA Terraform 1.10+ resource whose values are never written to state or to a plan file; provider 8 ships six, including google_secret_manager_secret_version.§26.11, §26.17
Lock file (.terraform.lock.hcl)The record of the exact provider versions and checksums resolved; a supply chain control, committed to version control, and rewritten by tofu init (§27.5).§26.1
Root moduleThe directory Terraform runs in; the only place a backend, a provider block, or variable values may appear.§26.6
State (Terraform)The address-to-resource map, attribute cache, and plaintext record of every value the provider read or wrote; treated as a secret store by §2.21.§26.13
Write-only argumentA provider argument whose value never reaches state or a plan file; provider 8 carries them on sixteen resources, and OpenTofu 1.11+ supports them too.§13.13, §26.17

G.7 Security Terminology §

The largest section, because security vocabulary spans every chapter: keys, secrets, audit evidence, perimeters, and compliance. A term here is defined once and owned by the section named beside it.

TermDefinition§
Access policyThe container for perimeters and access levels; one organization-level policy per organization, plus up to 50 folder- and project-scoped policies.§20.2
ACLAccess control list — the per-object Cloud Storage mechanism that uniform bucket-level access disables.§11.8
Crypto-shreddingDestroying a key version to render every ciphertext under it permanently unrecoverable; erases data only if every copy is under that version alone.§14.17
Customer-managed encryption keys (CMEK)Google-performed encryption under a key you administer and can revoke or destroy independently of the resource it protects.§14.8
Customer-supplied encryption keys (CSEK)Key material supplied per API request; Google never stores it, only a validation hash, so a lost key permanently loses the data.§14.9
Data encryption key (DEK)A locally generated key used to encrypt a payload directly in envelope encryption; wrapped by a KEK for storage.§14.16
DLPThe former brand for Sensitive Data Protection; the API and several commands still carry the name.§15.4
External Key Manager (EKM)Cloud KMS protection level where the key material lives in an external system (Fortanix, Futurex, Thales) reached over the internet or a VPC path; unreachability makes protected data unavailable.§14.11
HSMHardware security module — the Cloud KMS protection level with hardware-held key material.§14.10
Key Access Justifications (KAJ)An Assured-Workloads-only control attaching a justification code to every Google-system cryptographic operation on an enrolled key; production use requires GA status per consuming service.§14.14
Key encryption key (KEK)The Cloud KMS key that wraps a data encryption key in envelope encryption; it never leaves Cloud KMS.§14.16
Key ringA named, regional, undeletable grouping of keys with no configuration of its own; location and name are permanent.§14.2
Key versionThe Cloud KMS resource holding actual key material for one algorithm and protection level; ciphertext is bound to the version that produced it.§14.4
Protection levelWhere a Cloud KMS key's cryptographic operations execute: software, hsm, external, or external-vpc.§14.5
Public access prevention (PAP)A bucket setting, inherited or enforced, blocking any public binding regardless of mechanism.§11.10
Service perimeterA boundary around a set of projects, enforced by the Google API front end, that denies cross-boundary access to restricted services even when IAM allows it.§20.2
Signed URLA time-limited URL, up to 7 days, granting access to one object independent of IAM for its lifetime.§11.11
Uniform bucket-level access (UBLA)Bucket setting making IAM the sole access-control mechanism, disabling ACLs; irreversible after 90 consecutive days active.§11.9
Version destroy TTLA configured delay between a destroy call and irreversible payload deletion, giving an investigation window.§13.2
VPC accessible servicesThe complementary control restricting which Google APIs the networks inside a perimeter may call at all.§20.3

Sources §

Every definition here is a compression of the section named beside it, and those chapters carry the sources behind them. This appendix asserts no Google Cloud fact independently of the chapter that owns it; the selection method, the terms held in reserve, and the stale definitions corrected while building it are recorded in .build/verify/appG.md.