Practice Exams:

Microsoft AZ-104: Azure Front Door vs. Application Gateway

Cloud & Architecture

Azure Front Door and Application Gateway both route HTTP traffic, but they operate at different points in a request’s journey. The right choice depends on global reach, regional network boundaries, origin exposure, failure domains and the kind of traffic the application actually serves. An application with users across continents has different entry-point needs from a private regional workload. Origin reachability, web application filtering, health probes, and failover routes should decide whether a global front door or regional gateway is appropriate.

On this page
  1. Start with where the first trusted entry point belongs
  2. Understand routing and failover behavior
  3. Protect origins and regional backends
  4. Put WAF policy at the right inspection point
  5. Work out caching, performance and regional locality
  6. Avoid legacy SKUs and unsupported assumptions
  7. Choose one service, or justify a two-layer pattern

Start with where the first trusted entry point belongs

Consider a retailer serving customers on several continents. Requests need a global entry point capable of choosing a healthy region, reducing latency and protecting exposed web endpoints. A different application may serve one internal business unit from a single Azure region and require private network integration with a regional backend. Both applications use HTTPS, yet their ingress designs need not be the same. A Microsoft AZ-104 global-versus-regional entry-point choice depends on the client path, TLS handling, and whether edge or application routing is required.

Azure Front Door operates as a global edge application-delivery service. Application Gateway is a regional Layer 7 load balancer inside an Azure virtual network. Microsoft distinguishes these internet ingress patterns by routing scope and network placement. Front Door can deliver global HTTP/S routing and acceleration; Application Gateway can route requests to regional pools under regional networking and policy controls.

Draw the request path before selecting a SKU: client, DNS lookup, TLS termination, web application firewall (WAF) policy, routing decision, origin or backend connection and response. State which components are public and which must be privately reachable. An attractive global architecture is not justified if the application has only one internal region and does not need edge service capabilities.

Understand routing and failover behavior

Front Door can route requests among supported origin groups according to routing configuration and health evidence. That helps multi-region applications move new requests away from an unhealthy region. It does not automatically synchronize sessions, databases or application state. A platform service that shifts traffic to another region while the new region lacks current user records has failed at the application layer despite successful network routing.

Application Gateway selects backend targets inside its configured regional deployment and can provide path-based rules, host-based listeners, TLS termination and health probes. A regional failure requires a separate global steering or failover mechanism if the application must continue outside that region. Do not describe a single regional Application Gateway as a complete multi-region resilience solution.

For either product, validate what a health probe actually proves. A response to /health may show the web server is up while a database or payment API is unavailable. Choose health checks that reflect the dependencies used for routing decisions without making them themselves fragile. During a drill, record routing convergence, new requests, existing sessions and recovered business functions.

Protect origins and regional backends

A common Front Door risk is bypass: the origin remains directly reachable and accepts malicious or unintended traffic that never passed through the edge WAF. Restrict origins using supported networking and identity controls. The current Front Door security recommendations cover origin protection and tier-specific mechanisms. Private Link origin support and managed WAF features require an appropriate Front Door tier; do not assume Standard and Premium offer identical capabilities.

An Application Gateway resides in a VNet, which supports regional segmentation and backend access strategies. The VNet association does not alone make the frontend private; deliberately choose public or private frontend addressing and configure subnet, NSG and backend requirements according to the deployment. Backend authentication, certificates and workload authorization remain separate controls from gateway route rules.

If both services are deployed, treat their connection as a new trust boundary rather than an invisible hop. Which component validates the client? How is the forwarded origin host established? Which hop terminates TLS and re-encrypts traffic? Where are source IP information, headers and error responses retained? These questions matter when investigating unauthorized access or misrouted requests.

Put WAF policy at the right inspection point

Both services have Web Application Firewall options, but the supported policy architecture, placement and feature set differ. A global edge WAF can block harmful requests before they reach a region; a regional WAF can enforce app-specific policies close to the backend. It is sometimes useful to combine them, but duplicate rules can create confusing false positives or an inconsistent incident response path.

Start with threats relevant to the application: common injection patterns, malicious bots where supported, abusive traffic, HTTP protocol misuse, exposure of sensitive routes and geographic access assumptions. Apply managed and custom rules in a staged manner, review detections with application teams, and convert a rule to blocking only after false positives are understood. A passing TLS handshake does not imply the request was safe, and a WAF cannot replace proper application authorization.

Plan exception ownership. Developers often ask for a broad WAF exclusion when one field fails validation. A safe change record identifies the exact route, parameter, client use case, rule ID and expiration, with logs to confirm effect. A global exception for one application may expose unrelated origins; scope exceptions to the intended policy and service.

Work out caching, performance and regional locality

Front Door’s edge capabilities can be valuable for widely distributed audiences and cacheable objects. Performance depends on application cache headers, dynamic versus static response mix, origin proximity and TLS behavior. A globally distributed front end does not make an uncached, slow single-region database fast. Capture a representative client-to-application latency budget and identify the expensive segments.

Application Gateway focuses on regional application routing rather than being a global CDN. It can support local paths, backend pools and regional security processing where the application needs close integration with its virtual network. If all users sit near one Azure region, a regional design may meet the objective with less operational complexity.

Treat caching as data governance. Responses containing personal or tenant-specific information need correct cache directives and keys; a configuration error can expose one user’s content to another. Test personalized endpoints separately from anonymous public content and confirm purge and cache invalidation procedures before cutover.

Avoid legacy SKUs and unsupported assumptions

Azure Front Door (classic) is a retirement-path service, with Microsoft currently scheduling retirement for March 31, 2027. Modern planning should evaluate supported Standard or Premium tiers rather than recommending classic for new use. Application Gateway v1 retired on April 28, 2026, and Microsoft no longer supports v1 resources. An architecture discussion must identify the product generation before comparing price or feature names.

Do not copy an old checklist that treats all Front Door plans or all Application Gateway generations alike. Check current availability for Private Link origins, WAF managed rule sets, TLS configuration, diagnostics and upgrade/migration constraints. Certain networking and security features belong to a particular tier or configuration mode.

Migration risk is not just a SKU swap. Existing domain names, certificates, DNS TTLs, custom rules, backend health probes and origin restrictions must be mapped to the replacement. When changing ingress, measure actual HTTP response codes, authenticated user flows and security logs from both platforms. Rollback should be defined before DNS changes begin.

Choose one service, or justify a two-layer pattern

For a multi-region public web application needing edge acceleration and automatic regional traffic selection, Front Door is a natural starting point. For a single-region application needing VNet-specific backend routing or private-facing regional ingress, Application Gateway often fits better. A layered deployment can make sense where Front Door provides global entry and a regional Application Gateway applies specialized backend policy.

The layering should solve an explicit need. Otherwise it introduces two sets of certificates, probes, WAF policy, diagnostic logs, failure paths and billable services. A combined solution is especially difficult to troubleshoot if one layer reports a healthy origin while the second layer cannot reach its backend. Create a mapping of response headers, request identifiers and the responsible teams for each hop.

A useful validation exercise sends the same authorized test traffic through normal and failover paths, inspects the selected region and backend, tests that direct origin bypass is prevented, and confirms that a legitimate blocked request can be diagnosed without disabling the entire WAF. The choice is successful when the application stays reachable and the exposure boundaries remain enforceable—not merely when the Azure resources deploy successfully.

Continue learning

Related guides

Azure Load Balancer vs. Application GatewayChoose Layer 4 load balancing or Layer 7 application routing.Troubleshooting Azure VNets, Routes, and DNSDiagnose Azure route tables, NSG rules, and DNS answers for failed connections.Azure Private Endpoints and DNSVerify private endpoint DNS, routing, and service-side authorization.

Related Posts

• Microsoft AZ-104 Exam Guide: Skills, Labs, and Preparation

• How Difficult Is the Microsoft AZ-104 Exam?

• Microsoft AZ-104 vs. AZ-103: What Changed?

• Microsoft AZ-104: Troubleshoot VNets, Routes, and DNS

• Microsoft AZ-104: Azure DNS Troubleshooting

• Microsoft AZ-104: Azure Well-Architected Operations

• Microsoft AZ-104: Zones and Failure Domains

• Microsoft AZ-104: Azure Firewall Egress Design

• Microsoft AZ-104: Azure Backup and Recovery Design

• Microsoft AZ-104: ExpressRoute Resilience Patterns