Home › Studies › AZ-305 › Decision Guides

Decision Guides

Which service? Quick answers across all four domains.

One-screen answers to "which service?" questions. Each row links back to a topic page for the details.

Compute

NeedChoose
Full OS control, lift and shiftVirtual machines (scale sets Flexible for scale, zones for HA)
Web app or API, least managementApp Service
Event-driven code, pay per executionFunctions Flex Consumption (Windows needs the legacy Consumption plan)
Microservices, scale to zero, event scaling, no clusterContainer Apps
Kubernetes API, CRDs, HelmAKS
Single short-lived container job, no orchestrationContainer Instances
Large parallel or HPC jobsAzure Batch (Spot for restartable work, dedicated for strict deadlines or MPI)
Single-tenant, compliance, more than 30 plan instancesApp Service Environment v3

Messaging and events

NeedChoose
React to a state change (blob created)Event Grid
Telemetry or streaming at high volume, Kafka clientsEvent Hubs (Standard or higher)
Orders, ordering, transactions, dead-letter, duplicate detectionService Bus Standard or Premium
Cheap background queueStorage Queues
Messages up to 100 MB, JMS, geo-replicationService Bus Premium
Publish an API to partners with throttlingAPI Management

Load balancing and traffic

NeedChoose
Global HTTP(S), WAF, edge cachingFront Door
Global DNS routing, any protocolTraffic Manager
Regional L7: path routing, TLS offload, WAFApplication Gateway v2
Regional L4 TCP/UDPStandard Load Balancer
Global L4 with a static anycast IPCross-region Load Balancer
Insert third-party firewall NVAs transparentlyGateway Load Balancer

Relational databases

NeedChoose
New cloud app, single databaseAzure SQL Database (vCore; serverless if idle, Hyperscale beyond 4 TB)
SQL Agent, cross-database queries, CLR, instance-level featuresSQL Managed Instance
OS access, SSRS on the same serverSQL Server on Azure VMs
Many tenant databases, varying peaksElastic pool
Open-source enginesPostgreSQL or MySQL Flexible Server
Hide data from DBAsAlways Encrypted

Cosmos DB

NeedChoose
Steady trafficProvisioned throughput
Spiky traffic with an SLA and several regionsAutoscale
Sporadic traffic, one region, lowest costServerless (one-way conversion to provisioned)
Read your own writes (default)Session consistency
Always the latest writeStrong (single write region)
Tenants over 20 GB per logical partitionHierarchical partition keys

Storage

NeedChoose
Zone failure, one regionZRS
Regional disasterGRS or GZRS
Read the secondary before failoverRA-GRS or RA-GZRS
Rare access, instant readsCool or Cold
Compliance archiveArchive (hours to rehydrate)
WORM, even admins can't deleteImmutable storage, locked time-based policy
Cloud file shares with AD permissionsAzure Files SMB with identity-based auth
POSIX file shareAzure Files NFS (LRS or ZRS only)

Backup, DR and HA

NeedChoose
Survive a zone failureZones, ZRS, zone-redundant services
Fail VMs over to another region or zoneAzure Site Recovery
Restore deleted or corrupted dataAzure Backup (immutable vault + multi-user authorization against ransomware)
SQL regional DR, unchanged connection stringFailover group
SQL undo yesterday's mistakePoint-in-time restore
Keep SQL backups for yearsLong-term retention
Cheapest regional SQL DR, RPO about 1 hourGeo-restore

Hybrid connectivity

NeedChoose
Private, predictable, not over the internetExpressRoute
Encrypted, quick and cheapSite-to-site VPN (also a backup path)
Connect two on-prem sites over Microsoft's networkExpressRoute Global Reach
Other geopolitical regionsExpressRoute Premium
Survive loss of a peering locationExpressRoute Metro
Hub with many regions and branchesVirtual WAN

Identity and keys

NeedChoose
Cloud sign-in that survives on-prem outagePassword hash sync
Enforce on-prem policy at sign-inPass-through authentication
Disconnected forests, multi-agent failoverEntra Cloud Sync
Customer sign-up appEntra External ID (external tenant)
Partners in your tenantB2B collaboration + entitlement management
Secrets without credentialsManaged identity + Key Vault (RBAC)
Single-tenant HSMManaged HSM or Azure Cloud HSM (Dedicated HSM is retiring)
Just-in-time adminPIM eligible assignment (P2)

Governance and monitoring

NeedChoose
What can be deployedAzure Policy
Who can do itRBAC (ABAC to narrow by attributes)
Stop deletion or changeResource lock (control plane only)
Fix existing non-complianceRemediation task
VM guest logsAzure Monitor Agent + DCR
Resource logs to a SIEMDiagnostic setting to Event Hubs
Alert when a resource is deletedActivity log alert
Silence alerts during maintenanceAlert processing rule