Reference · Glossary

AZ-900 Glossary

Precise definitions for terms that show up across lessons — especially the ones that sound similar but aren't. Grows as lessons are added; every lesson should use these terms consistently with what's defined here.

Cloud concepts (Lessons 1–2)

Shared responsibility model

The split of security/operational duties between Microsoft and the customer. Microsoft always owns physical datacenter security; everything above that (OS, app, data) shifts depending on IaaS/PaaS/SaaS. See Lesson 1.

Deployment models

Public cloud — shared infrastructure owned by the provider. Private cloud — dedicated infrastructure for one organization. Hybrid cloud — a connected mix of both. See Lesson 1.

Consumption-based (pricing) model

Pay only for what you use (OpEx), instead of buying hardware upfront sized for peak load (CapEx). See Lesson 1.

Serverless

You deploy code; the platform provisions, patches, and scales the compute automatically — including down to zero. Still runs on servers, you just never manage them. See Lesson 1.

⚠ Confusable pair: vertical vs. horizontal scaling

Vertical scaling (scale up/down) changes the capability of one resource — e.g. more CPU/RAM on a VM. Horizontal scaling (scale out/in) changes the number of resource instances — e.g. adding more VMs behind a load balancer. See Lesson 2.

High availability

A system stays up despite disruptions, backed by a per-service SLA (e.g. 99.9%, 99.99% uptime guarantees). See Lesson 2.

Reliability

A system's ability to recover from failure and keep functioning — supported by decentralizing across Azure regions. A pillar of the Azure Well-Architected Framework. See Lesson 2.

Predictability

Performance predictability — autoscaling/load balancing keep the experience consistent under shifting demand. Cost predictability — real-time usage tracking and the Pricing Calculator let you forecast spend. See Lesson 2.

⚠ Confusable pair: management of vs. management in the cloud

Management of the cloud — automated things that happen to your resources, no hands involved: auto-scale, deployment templates, health monitoring with auto-replace, metric alerts.

Management in the cloud — the interfaces you personally use to reach in and act: web portal, CLI, APIs, PowerShell.

Mnemonic: "in" = interfaces you use by hand. "of" = automated, happens on its own. Flagged as a genuine trouble spot on first pass — see learning record 0001 if this one isn't sticking; ask your teacher rather than re-reading silently. See Lesson 2.

IaaS / PaaS / SaaS

Infrastructure as a Service — you manage OS, patching, and config; provider manages hardware/connectivity/physical security. Platform as a Service — provider also manages OS, middleware, and runtimes; you manage only your app and data. Software as a Service — provider manages almost everything; you manage data, identity, and access. Control decreases and convenience increases as you move IaaS → PaaS → SaaS. See Lesson 3.

Azure architecture (Lesson 4)

Region / availability zone / region pair

Region — a geographic area with one or more networked datacenters. Availability zone — a physically separate datacenter within a region (minimum 3 per AZ-enabled region), an isolation boundary. Region pair — two regions in the same geography, ≥300 miles apart, for disaster resilience. See Lesson 4.

Resource / resource group / subscription / management group

Nesting order, top to bottom: management group (governance at scale, up to 6 levels deep) → subscription (billing + access-control boundary) → resource group (a resource belongs to exactly one, at a time) → resource (the actual thing deployed). Policy and access inherit downward. See Lesson 4.

Compute types: VM / container / function

Virtual machine — full OS control, IaaS, slowest start, billed while provisioned. Container — shares host OS, lightweight, fast start, billed while running. Function — serverless, wakes on an event, billed only for execution time. See Lesson 5.

⚠ Confusable pair: scale sets vs. availability sets

Scale sets — identical, load-balanced VMs that auto-scale out/in with demand (elastic capacity). Availability sets — VMs grouped by fault domain + update domain so one failure/maintenance event can't take them all down (resiliency, not scale). See Lesson 5.

⚠ Confusable pair: VPN Gateway vs. ExpressRoute

VPN Gateway — encrypted tunnel over the public internet (site-to-site, point-to-site, or VNet-to-VNet). ExpressRoute — dedicated private connection via a provider that never touches the public internet; higher reliability/bandwidth, used for compliance-sensitive or latency-critical workloads. Even over ExpressRoute, DNS queries and CDN requests still go over the public internet. See Lesson 6.

Public vs. private endpoint

Public endpoint — public IP, reachable from anywhere. Private endpoint — private IP inside a VNet, reachable only from within. See Lesson 6.

VNet peering

Direct private connection between two virtual networks (even across regions), over the Microsoft backbone, never the public internet. See Lesson 6.

Azure storage (Lessons 7–8)

Blob access tiers

Hot — frequent access. Cool — infrequent, 30-day minimum. Cold — infrequent, 90-day minimum. Archive — rare access, 180-day minimum, highest rehydration latency, lowest storage cost. See Lesson 7.

⚠ Confusable set: LRS / ZRS / GRS / GZRS

Azure Storage always keeps 3 copies in the primary region. LRS — 3 copies, one datacenter. ZRS — 3 copies across availability zones. GRS — LRS in primary + async LRS copy in a secondary region. GZRS — ZRS in primary + async LRS copy in a secondary region. Add RA- (RA-GRS/RA-GZRS) to read the secondary copy before failover. See Lesson 7.

AzCopy / Storage Explorer / Azure File Sync

AzCopy and Storage Explorer (GUI wrapper around AzCopy) are one-directional, on-demand file copy tools. Azure File Sync is a standing, bi-directional sync between Azure Files and a local Windows Server cache. See Lesson 8.

Azure Migrate vs. Azure Data Box

Azure Migrate — hub for online, real-time assessment/migration of on-prem infrastructure. Azure Data Box — physical device (up to 80 TB) for offline bulk transfer when bandwidth is the bottleneck. See Lesson 8.

Identity & access (Lesson 9)

Microsoft Entra ID vs. Microsoft Entra Domain Services

Entra ID — cloud identity/access management (auth, SSO, app/device management). Entra Domain Services — managed classic domain services (domain join, group policy, LDAP, Kerberos/NTLM) for legacy apps, no domain controllers to maintain yourself. See Lesson 9.

SSO / MFA / passwordless

SSO — one sign-in, many apps. MFA — a second factor (know/have/are). Passwordless — no password at all (Windows Hello for Business, Microsoft Authenticator app, FIDO2 keys); rated highest on both security and convenience. See Lesson 9.

⚠ Confusable pair: RBAC vs. Conditional Access

RBAC — authorization: what you're allowed to do, scoped to a resource/group/subscription/management group, inherits downward. Conditional Access — a sign-in-time gate: signals (who/where/device) → decision → enforcement (allow/block/require MFA), independent of RBAC permissions. See Lesson 9.

B2B collaboration vs. B2B direct connect

B2B collaboration — external partner becomes a guest user in your directory. B2B direct connect — mutual trust between tenants, no directory presence (currently Teams shared channels). See Lesson 9.

Security models (Lesson 10)

Zero Trust

Assumes breach and verifies every request explicitly, regardless of network location. Three principles: verify explicitly, use least privilege access, assume breach. See Lesson 10.

⚠ Confusable pair: Zero Trust vs. defense-in-depth

Zero Trust — a mindset for access decisions: verify every request explicitly, no matter where it comes from. Defense-in-depth — a structure of nested layers (physical → identity & access → perimeter → network → compute → application → data) around the environment, so one breach doesn't mean total loss. Complementary: Zero Trust governs who passes each layer; defense-in-depth is the set of layers.

Layer order as a castle: physical (walls) → identity & access (gate guards) → perimeter (moat — stops mass attacks like DDoS before they arrive) → network (roads/checkpoints between things already inside — segmentation) → compute (buildings) → application (what happens inside them) → data (the vault, always innermost). Flagged as a trouble spot — perimeter vs. network especially — see learning record 0002. See Lesson 10 · Lesson 11.

Microsoft Defender for Cloud

Security posture management and threat protection service, native to Azure and extendable to hybrid/multicloud via Azure Arc. Three pillars, best read as a timeline: continuously assess ("what's wrong right now?" — raw vulnerability findings) → secure ("how do I improve, and track it?" — benchmark-based recommendations rolled into the secure score) → defend ("am I under attack right now?" — alerts, kill-chain correlation, advanced threat protection). The secure score belongs to Secure, not Assess — see learning record 0002. See Lesson 10 · Lesson 11.

Cost management (Lesson 13)

⚠ Confusable pair: Pricing calculator vs. TCO calculator

Pricing calculator — estimates the cost of a new Azure deployment planned from scratch. TCO calculator — compares existing on-premises infrastructure cost against running that same workload in Azure. Pricing calculator looks forward from nothing; TCO calculator looks sideways from what you already have. See Lesson 13.

Reservations vs. Spot pricing

Reservation — discount for committing to 1 or 3 years upfront. Spot pricing — deep discount on unused spare capacity, can be evicted anytime, unrelated to any commitment length. See Lesson 13.

Tags

Name/value metadata attached to a resource, resource group, or subscription. Doesn't change how anything runs — lets you group and report on resources (e.g. in Cost Management) across resource-group boundaries. A resource sits in exactly one resource group but can carry many tags. See Lesson 13.

Governance & compliance (Lesson 14)

⚠ Confusable pair: Azure Policy vs. Azure RBAC

RBAC — controls who can perform which actions, scoped to a role assignment. Azure Policy — controls what a resource's properties are allowed to be, regardless of who's creating it; can deny a request even from a user with full RBAC write permission. RBAC gates the actor; Policy gates the outcome. See Lesson 14.

Resource locks: CanNotDelete vs. ReadOnly

CanNotDelete — blocks deletion only, resource can still be modified. ReadOnly — blocks both modification and deletion. Both override RBAC entirely, including Owner. Locks inherit downward (subscription → resource group → resource); the more restrictive lock wins. See Lesson 14.

Microsoft Purview

Unified data governance across on-premises, multicloud, and SaaS — not Azure-only. Discovers, catalogs, and classifies data, then layers on risk and compliance management. See Lesson 14.

Managing & deploying resources (Lesson 15)

⚠ Confusable pair: declarative (ARM templates) vs. imperative (CLI/PowerShell scripts)

Declarative — state the desired end result; the platform works out how to get there (ARM templates, Bicep). Imperative — specify the exact sequence of steps to run, in order (a CLI or PowerShell script). Declarative says "what," imperative says "how, in order." See Lesson 15.

Azure Arc

Projects non-Azure resources (on-premises, other clouds, Kubernetes anywhere) into Azure Resource Manager, so Policy, tags, and RBAC apply to them the same as native Azure resources — one control plane for hybrid/multicloud. See Lesson 15.

Portal / Cloud Shell / CLI / PowerShell / mobile app

Interchangeable doors into the same Azure Resource Manager. Portal — web GUI. Cloud Shell — browser-based CLI, no install, persistent file share. Azure CLI — locally installed, cross-platform, az commands. Azure PowerShell — locally installed, cmdlet syntax. Mobile app — monitor and act from a phone. See Lesson 15.

Monitoring tools (Lesson 16)

⚠ Confusable pair: Azure Advisor vs. Azure Service Health

Advisor — personalized recommendations about your own resources (reliability, security, performance, cost, operational excellence) — things you can fix. Service Health — status of the Azure platform itself (outages, planned maintenance, advisories), filtered to services/regions you use — things Microsoft has to fix. Advisor looks at you; Service Health looks at Azure. See Lesson 16.

Azure Monitor: Metrics vs. Log Analytics vs. Application Insights

Metrics — lightweight numerical time-series, near real-time. Log Analytics — richer structured log data queried with KQL. Application Insights — Azure Monitor's APM feature, tracking requests/exceptions/dependencies for a specific running app rather than raw infrastructure. See Lesson 16.

← Back to mission · Roadmap