Study guide

5.2 Database Service & Compute Tier Design

Exam AZ-305: Designing Microsoft Azure Infrastructure Solutions Exam

Track your progress as you read

Finished sections get a check mark.

1 Introduction & Exam Orientation

1.1 AZ-305 Current Exam Facts & Format 1.2 Candidate Profile, Prerequisites & the AZ-104 Requirement 1.3 Skills Outline, Domain Weights & How to Use This Guide

2 Chapter 2: Logging, Monitoring & Diagnostics Design

2.1 Logging Solution Design 2.2 Log Routing Design 2.3 Azure Monitor Metrics & Alerting Design 2.4 Application Insights & Diagnostic Monitoring Design

3 Chapter 3: Authentication & Authorization Design

3.1 Authentication Solution Design 3.2 Identity Management Design 3.3 Authorization for Azure Resources 3.4 Authorization for On-Premises Resources 3.5 Secrets, Certificates & Key Management Design

4 Chapter 4: Governance, Compliance & Identity Governance Design

4.1 Management Group & Subscription Hierarchy Design 4.2 Resource Group Organization & Tagging Strategy Design 4.3 Compliance Management Design 4.4 Identity Governance Design

5 Chapter 5: Relational Data Storage Design

5.1 Relational Data Storage Selection 5.2 Database Service & Compute Tier Design 5.3 Database Scalability Design 5.4 Relational Data Protection Design

6 Chapter 6: Semi-Structured & Unstructured Data Storage Design

6.1 Semi-Structured Data Storage Design 6.2 Unstructured Data Storage Design 6.3 Storage Tier & Cost-Performance Trade-off Design 6.4 Unstructured Data Protection & Durability Design

7 Chapter 7: Data Integration & Analytics Design

7.1 Data Integration Pipeline Design 7.2 Event-Based & Real-Time Data Integration Design 7.3 Data Analytics Platform Design

8 Chapter 8: Backup & Disaster Recovery Design

8.1 Recovery Objectives Design: Calculating and Meeting RTO/RPO 8.2 Azure Site Recovery Architecture for Hybrid Workloads 8.3 Compute Backup & Recovery Design 8.4 Database Backup & Recovery Design 8.5 Unstructured Data Backup & Recovery Design

9 Chapter 9: High Availability Design

9.1 Compute High Availability: Availability Sets & Availability Zones 9.2 Compute High Availability: Load Balancing & VM Scale Set Design 9.3 Relational Data High Availability Design 9.4 Semi-Structured & Unstructured Data High Availability Design

10 Chapter 10: Compute Solution Design

10.1 Compute Requirements Analysis & Virtual Machine-Based Solution Design 10.2 Container-Based Solution Design (AKS, Container Apps, Azure Container Instances) 10.3 Serverless-Based Solution Design (Azure Functions, Logic Apps) 10.4 Batch Processing Solution Design (Azure Batch, HPC Workloads)

11 Chapter 11: Application Architecture Design

11.1 Messaging Architecture Design (Service Bus, Queue Storage) 11.2 Event-Driven Architecture Design (Event Grid, Event Hub) 11.3 API Integration & Application Caching Design (API Management, Azure Cache for Redis, CDN) 11.4 Application Configuration Management & Automated Deployment Design (App Configuration, DevOps Pipelines, Deployment Slots)

12 Chapter 12: Migration Design

12.1 Migration Assessment: Cloud Adoption Framework & On-Premises Evaluation 12.2 IaaS & PaaS Workload Migration Design 12.3 Database Migration Design (Data Migration Service, Database Migration Assistant) 12.4 Unstructured Data Migration Design (AzCopy, Data Box, Storage Migration Service)

13 Chapter 13: Network Solution Design

13.1 Internet Connectivity Design 13.2 Hybrid Connectivity Design 13.3 Network Security Optimization Design 13.4 Network Performance Optimization & Load-Balancing/Routing Design

14 Chapter 14: Final Review, Exam Strategy & Career Path

14.1 Cross-Domain Scenario Practice: Combining Identity, Data, Continuity & Infrastructure Decisions 14.2 Common AZ-305 Decision Traps 14.3 Final Study Plan & Practice Assessment Strategy 14.4 After AZ-305: Retake Policy, Renewal & Career Path

5.2 Database Service & Compute Tier Design

Key Takeaways Explain This Topic AI

vCore is Microsoft's recommended purchasing model for new deployments because it exposes compute/storage separately and is the only model that supports Azure Hybrid Benefit.

General Purpose uses remote storage for balanced cost; Business Critical uses local SSD with a built-in Always On availability group for lowest latency and one free readable secondary; Hyperscale decouples storage from compute for databases up to 128 TB with near-instant backups.

Serverless is a compute option inside General Purpose only — it autoscales, auto-pauses during inactivity, and bills per second, fitting intermittent workloads that can tolerate a brief resume delay.

DTU bundles compute, memory, and I/O into fixed Basic/Standard/Premium tiers and cannot use Azure Hybrid Benefit or reach Business Critical/Hyperscale tiers.

Match the tier to the defining constraint: latency and built-in HA point to Business Critical; size and fast backup/restore point to Hyperscale; cost-sensitive intermittent use points to serverless General Purpose.

Last updated: July 2026

Why This Topic Matters

Once you have picked a relational engine (5.1), AZ-305 tests the next design decision under the same official bullet family: "Recommend a database service tier and compute tier." This is a right-sizing and cost/performance trade-off question — Azure SQL Database and Managed Instance both offer more than one purchasing model and more than one service tier , and the exam expects you to match workload characteristics (predictability, latency sensitivity, size, budget) to the correct combination.

Purchasing Models: DTU vs. vCore

Model How it works Best for

DTU (Database Transaction Unit) A blended measure of CPU, memory, and I/O bundled into preset tiers ( Basic , Standard , Premium ) with fixed storage per tier Simple, predictable workloads where you do not want to reason about compute and storage separately; smaller/legacy deployments already sized in DTUs

vCore You independently choose the number of virtual cores, the hardware generation, and the storage amount Workloads that need transparent, granular control over compute vs. storage cost, want Azure Hybrid Benefit (reusing existing SQL Server licenses for a discount), or need service tiers not available under DTU (Business Critical, Hyperscale)

Key fact: vCore is Microsoft's recommended model for new deployments because it exposes hardware and licensing transparently and unlocks Azure Hybrid Benefit , which the DTU model cannot use. On the exam, a scenario that mentions "the organization has existing SQL Server licenses with Software Assurance and wants to reduce Azure costs" is a strong signal to recommend the vCore model with Azure Hybrid Benefit , not DTU.

vCore Service Tiers

Tier Storage architecture Characteristics Best for

General Purpose Remote (network) storage, single compute node Balanced compute/storage cost, 99.99% SLA Most business workloads without extreme latency or built-in read-replica needs

Business Critical Local SSD storage, built-in Always On availability group across replicas Lowest latency, highest resilience, includes one free readable secondary replica for read scale-out Mission-critical, latency-sensitive OLTP (e.g., trading, ticketing)

Hyperscale Decoupled storage and compute via a scale-out page-server architecture Storage autoscaling up to 128 TB, near-instant backups and fast restores independent of database size, up to 30 named replicas with independently sized compute Fast-growing databases, large analytical or mixed workloads, scenarios needing rapid backup/restore regardless of size

The Serverless Compute Tier

Serverless is a compute option available only in the vCore General Purpose tier (not a separate service tier). Its defining behaviors:

Autoscaling : compute automatically scales between a configured minimum and maximum vCore count based on workload demand.

Auto-pause : after a configurable period of inactivity (the auto-pause delay), the database pauses completely — only storage is billed while paused.

Per-second billing : you pay for vCores consumed and memory, billed per second of active compute, rather than for a fixed provisioned size around the clock.

Automatic resume : the first new connection triggers a resume, which takes several seconds to a couple of minutes depending on cache state.

Serverless is the correct recommendation for workloads with intermittent, unpredictable usage and tolerance for a brief resume delay — classic examples are dev/test databases, low-traffic line-of-business apps used only during business hours, and new applications where usage patterns are not yet known. It is the wrong choice for workloads that cannot tolerate any cold-start latency (e.g., a customer-facing checkout database that must always be warm) — for those, use provisioned compute in the appropriate tier instead.

Exam Scenarios

Scenario A: A finance team runs a batch reporting database used only 9am-5pm on weekdays; it sits idle every night and weekend. Cost efficiency matters more than instant availability. Answer: vCore General Purpose tier with serverless compute and an appropriate auto-pause delay.

Scenario B: A stock-trading platform requires sub-millisecond storage latency and automatic failover with zero data loss if the primary node fails. Answer: vCore Business Critical tier — its local SSD storage and built-in Always On availability group deliver both requirements.

Scenario C: An analytics database is growing rapidly toward multiple terabytes, and the team needs backups and restores to remain fast regardless of size, plus the ability to add dedicated read replicas for reporting without affecting the primary. Answer: Hyperscale — its decoupled storage/compute design keeps backup/restore times roughly constant as data grows, and named replicas isolate reporting load.

Scenario D: An organization already owns SQL Server Enterprise Edition licenses with active Software Assurance and wants to minimize Azure spend on a new deployment. Answer: vCore purchasing model with Azure Hybrid Benefit applied — impossible under the DTU model.

Common Traps

Recommending DTU just because a scenario is "simple" — if Azure Hybrid Benefit, Hyperscale, or Business Critical is mentioned or implied, vCore is required.

Confusing serverless with a separate service tier — it is a compute option within General Purpose , not its own tier, and it is not available in Business Critical or Hyperscale.

Picking Business Critical purely for "high availability" when Hyperscale is a better fit for a workload whose defining constraint is size and backup/restore speed rather than sub-millisecond latency.