Microsoft AZ-305: Integration and Messaging Architecture — Study Guide

By Laurynas Brazys · 11 August 2026 · #microsoft #az-305 #certification #exam-prep #study-guide

Part of the Microsoft Azure Solutions Architect Expert AZ-305 — Study Guide. Practice with verified answers in the Microsoft exam hub, or take timed practice tests on ExamRoll.io.

Overview

Integration and messaging architecture on Azure hinges on selecting the correct service for command, event, and data movement semantics; designing for reliability, ordering, and scale; and integrating hybrid systems securely. Core building blocks include Azure Service Bus for enterprise messaging with rich brokers, Azure Event Grid for reactive event routing, Azure Event Hubs for high-throughput streaming ingestion, and Storage Queues for simple queueing. Surrounding these are Azure Logic Apps for process automation, API Management for governance and developer experience, Azure Data Factory for ETL/ELT, Azure Relay for firewall-friendly on-premises connectivity, and Azure Notification Hubs for mobile push. Sound design uses the right service for the job, models contracts and failure modes explicitly, and applies orchestration or choreography patterns where appropriate.

Messaging and Eventing on Azure

Azure Service Bus is the enterprise broker for commands and workflows that require ordered delivery, transactions, FIFO within groups, and dead-letter handling. Queues deliver point-to-point messaging between one producer and one competing consumer group. Topics enable pub/sub with multiple independent subscriptions that can filter and route messages using SQL-like filters and actions. Message sessions group related messages under a sessionId, enabling FIFO and stateful processing per group; the session lock ensures a single consumer processes a session at a time. Dead-letter queues capture poison messages when max delivery count is exceeded, TTL expires, or when explicitly dead-lettered, enabling quarantine and later inspection; each queue or subscription has a $DeadLetterQueue subqueue. Transactions allow atomic send/receive/complete operations across entities within the same namespace, ensuring, for example, that a message is completed only if follow-up messages are successfully sent (send-via supports cross-entity workflows).

Azure Event Grid is a fully managed event router for reactive, push-based patterns. It uses either the native Event Grid schema or CloudEvents 1.0. Event subscriptions target endpoints such as Functions, Logic Apps, Service Bus, Event Hubs, WebHooks, and Storage Queues, with subject and advanced filters to reduce noise. Delivery is retried with exponential backoff; subscriptions support a configurable retry policy (maximum delivery attempts and event time to live) and can dead-letter to a Storage account for undeliverable events. Handshake validation secures WebHook endpoints, and managed identities simplify publishing and delivery to Azure endpoints.

Azure Event Hubs ingests telemetry and streaming data at massive scale with partitioned logs. Partitions provide parallelism and ordering within a partition; choose a partition key to keep related events ordered. Partition count determines scale and cannot be reduced after creation, so size for future throughput. Consumer groups provide independent views of the event stream for different applications without interfering with each other’s offsets. Capture persistently offloads data to Azure Blob Storage or Data Lake Storage in near real time based on time/size windows, typically in Avro format, enabling batch analytics without impacting ingestion. The built-in Schema Registry stores Avro/JSON schemas with versioning and compatibility policies, enabling producers and consumers to validate and evolve contracts safely.

Storage Queues offer simple, cost-effective at-least-once delivery with invisibility timeouts and basic dead-lettering via message TTL and application-managed poison handling. They lack transactions, sessions, and advanced routing, but excel for basic decoupling and large fan-out at low cost.

When to choose which:

• Use Service Bus for commands, workflows, and integration scenarios requiring FIFO (via sessions), transactions, deferral, duplicate detection, and dead-letter auditing.
• Use Event Grid for lightweight, push-based, fan-out notifications from Azure services or custom apps with fine-grained filtering and near-real-time reactions.
• Use Event Hubs for high-throughput streaming telemetry and log ingestion, with independent consumers and downstream analytics.
• Use Storage Queues for simple producer/consumer decoupling where advanced broker features are unnecessary and cost simplicity is paramount.

Integration, APIs, and Hybrid Connectivity

Azure Logic Apps provides managed workflow automation with hundreds of connectors. In Consumption (multi-tenant), you pay per action with automatic scaling and multi-tenant connectors; ideal for bursty use. Standard (single-tenant) runs on the Functions runtime with stateful/stateless workflows, higher throughput, custom connectors and built-in connectors running in-process, containerization, local development, and VNET/private endpoint integration; ideal for enterprise isolation and predictable capacity. The Integration Service Environment (ISE) is a legacy dedicated stamp for private networking and data locality; new designs generally prefer Logic Apps Standard with VNET integration or deployment to App Service Environment v3.

Azure API Management (APIM) provides an abstraction and governance layer over APIs. Policies apply at inbound, backend, and outbound stages to enforce cross-cutting concerns such as validate-jwt, rate-limit-by-key, quota, set-header, retry, cache-lookup/store, and set-backend-service for dynamic routing. Products group one or more APIs, bundle policy behavior, and are publishable to specific groups. Subscriptions issue keys per consumer or per product to meter and control access; keys can be rotated and tied to quotas. The developer portal enables self-service discovery, documentation, try-it, and onboarding workflows, while the self-hosted gateway allows hybrid control plane/edge deployments for on-prem or other clouds.

Azure Relay enables inbound connectivity to on-prem services without opening inbound firewall ports. Hybrid Connections uses WebSockets over TLS 443 for general-purpose, bidirectional socket communication initiated outbound from on-prem and from clients to the Relay, useful for HTTP and arbitrary protocols encapsulated over WebSockets. WCF Relay exposes on-prem WCF endpoints (NetTcp/HTTP) through the relay with transport- or message-level security and claims-based access; it is ideal for existing WCF services needing secure, firewall-friendly exposure.

Azure Notification Hubs is a cross-platform push broker that abstracts platform notification systems (APNs for iOS, FCM for Android, WNS for Windows, ADM for Amazon). Backends register devices or installations with tags and templates to target and personalize notifications at scale. Platform credentials are required per PNS: APNs p8 key or certificate, FCM server key/credentials, WNS package SID/secret. Notification Hubs handles fan-out, throttling, and token management so application code remains PNS-agnostic.

Data Movement and Streaming Analytics

Azure Data Factory (ADF) orchestrates data integration across hybrid estates. Integration runtimes (IRs) host compute for activities: Azure IR for serverless copy and data flows within Azure, Self-hosted IR for data movement/compute within private networks without opening inbound ports, and Azure-SSIS IR to lift-and-shift SSIS packages to managed clusters. Pipelines orchestrate activities with control flow (dependencies, loops, branches, triggers) and parameterization for reuse. Mapping Data Flows provide code-free, Spark-based transformations at scale with schema drift handling and partitioning controls; use when transformations are complex but you want managed compute. Linked services define connection metadata and credentials for sources, sinks, and compute; datasets and data flow sources/sinks reference these, enabling secure reuse and RBAC.

Event Hubs integrates with analytics through Capture to persistent storage, then Azure Synapse or Databricks can process Avro files in micro-batches. Schema Registry simplifies deserialization and evolution in streaming jobs by centralizing contracts, avoiding brittle, implicit typing across producers and consumers.

Reliability and Event-Driven Patterns

Designing for reliability begins with explicit failure handling. Service Bus provides dead-letter queues per entity; consumers should monitor and triage DLQs, optionally auto-forwarding to analysis queues. Use duplicate detection and idempotent handlers to avoid double-processing. Leverage transactions to atomically settle receives and send outbox messages, and use sessions for ordered processing per business entity while scaling across sessions. Event Grid’s retry policy uses exponential backoff with configurable limits; configure dead-lettering to Storage for auditability and build replay tooling. Event Hubs guarantees at-least-once delivery; checkpointing via SDKs (or Azure Functions triggers) ensures progress tracking per partition/consumer group. Storage Queues rely on visibility timeouts and poison message patterns you implement explicitly.

Event-driven architecture typically uses choreography or orchestration. Choreography distributes coordination across services reacting to events from peers. It is loosely coupled, scalable, and resilient to partial failures but can become hard to visualize and govern, and compensations are dispersed. Orchestration centralizes flow control in an orchestrator such as Azure Durable Functions, Logic Apps, or a workflow engine, improving observability, timeout/compensation logic, and human-in-the-loop steps at the cost of tighter coupling to the orchestrator. The saga pattern implements long-lived, multi-step transactions with compensating actions instead of two-phase commits. In choreography, each service listens for domain events and emits compensations as needed; in orchestration, the orchestrator invokes activities and triggers compensations on failure or timeout. On Azure, implement sagas using Durable Functions (stateful orchestration, retries, timeouts, compensation patterns) with Service Bus for reliable command delivery, or with Logic Apps Standard for robust enterprise workflows and built-in connectors.

Practical Problem Scenario

Contoso Retail modernizes its order processing across an on-prem ERP, an e-commerce site, mobile apps, and downstream analytics. The solution must support partner notifications, real-time telemetry analysis, secure on-prem access, and mobile push, with strict ordering and compensations for payment and inventory.

1. Ingest and command workflow

• Use Azure Service Bus topics for order commands. Subscriptions segment processing (Payments, Inventory, Shipping). Enable message sessions keyed by OrderId to guarantee per-order FIFO and single-concurrency. Rationale: Service Bus provides sessions, transactions, and dead-lettering required for ordered, reliable business workflows.

2. Orchestration and compensations

• Implement a saga with Azure Durable Functions. Activities call payment gateway, reserve inventory, and create shipment; compensations refund or restock on failure. Use Service Bus triggers and outputs inside transactional scopes to atomically complete incoming messages and publish follow-ups. Rationale: Centralized orchestration simplifies timeouts, retries, and compensations while preserving broker reliability.

3. Event-driven notifications

• Publish domain events (OrderPlaced, OrderShipped) to Azure Event Grid custom topics. Partners and internal apps subscribe with filters by subject prefix. Configure dead-lettering to a Storage account and a retry policy with bounded attempts. Rationale: Event Grid offers low-latency push with fine-grained filtering and dead-letter audit for broad fan-out.

4. Telemetry streaming and analytics

• Send clickstream and app telemetry to Azure Event Hubs with 8 partitions keyed by user session. Enable Capture to Data Lake Storage every 5 minutes or 100 MB. Register Avro schemas in Schema Registry and enforce schema compatibility. Rationale: Event Hubs scales ingestion independently of consumers; Capture decouples analytics; Schema Registry maintains contract discipline.

5. Data integration

• Use Azure Data Factory with a Self-hosted Integration Runtime on-prem to extract ERP data securely, and an Azure IR to land curated data into Synapse. Build pipelines and mapping data flows for SCD processing and enrichment with Event Hubs Capture files. Rationale: ADF orchestrates hybrid data movement and transformations with managed compute and governed connections via linked services.

6. APIs and partner governance

• Front all partner-facing endpoints with Azure API Management. Expose order status and webhook registration APIs. Apply policies: validate-jwt for partner-issued tokens, rate-limit-by-key for stricter partner throttles, rewrite/set-backend-service to route to existing Logic Apps without code changes. Publish a product for partners requiring subscriptions and keys; onboard via the developer portal. Rationale: APIM enforces security, throttling, and provides self-service onboarding without touching backend logic.

7. Workflow automation and connectors

• Use Azure Logic Apps Standard for back-office automation (e.g., send emails, update Dynamics) with built-in connectors and VNET integration. Rationale: Single-tenant performance, private connectivity, and enterprise-grade connectors simplify integration with SaaS and line-of-business apps.

8. Hybrid connectivity

• Expose selective on-prem ERP services via Azure Relay WCF Relay for existing WCF endpoints and Hybrid Connections for lightweight HTTP/WebSocket applications, avoiding inbound firewall changes. Rationale: Relay provides secure, outbound-initiated connectivity without VPN complexity.

9. Mobile push

• Send shipping updates via Azure Notification Hubs using tags for device/user segmentation and templates for localization. Configure APNs token credentials and FCM keys. Rationale: Notification Hubs abstracts PNS differences and scales push delivery.

This design maps each requirement to purpose-built services: Service Bus for reliable commands, Durable Functions for orchestrated sagas, Event Grid for push notifications, Event Hubs + Capture + Schema Registry for streaming analytics, ADF for hybrid ETL/ELT, APIM for governance, Logic Apps for enterprise automation, Relay for on-prem access, and Notification Hubs for mobile engagement.

← Security Architecture and Zero Trust · All domains · Monitoring →

Practice these questions → · Timed practice on ExamRoll.io →

Browse Microsoft →