Traffic Manager: DNS-Based Routing

Traffic Manager works at the DNS layer. When a client queries a Traffic Manager FQDN, it returns the IP of the best endpoint based on the routing method. Traffic Manager doesn’t proxy traffic — it redirects at the DNS level.

🎬 Video coming soon

Traffic Manager Routing Methods

Traffic Manager Routing Methods

~12:00

The Six Routing Methods

Method	How It Works	Best For
Priority	Routes to the highest-priority (lowest number) healthy endpoint. Failover to next if primary is down.	Active-passive DR. Primary in one region, backup in another.
Weighted	Distributes traffic based on weights (1-1000). Weight 300 gets 3x more traffic than weight 100.	Gradual migration, canary deployments, capacity-based distribution.
Performance	Returns the endpoint with lowest latency from the client’s DNS resolver location.	Multi-region apps where users should reach the closest region.
Geographic	Routes based on the geographic origin of the DNS query. Users in a region always go to the assigned endpoint.	Data sovereignty, regional content, compliance requirements.
Multivalue	Returns multiple healthy endpoint IPs in a single DNS response. Client chooses.	Simple redundancy — client-side failover. Only works with external endpoints with IPs.
Subnet	Maps specific client IP ranges to specific endpoints.	Testing, VIP routing, per-customer endpoints.

Traffic Manager Routing Methods Comparison

Method	How It Works	Best For
Priority	Routes to the highest-priority healthy endpoint; fails over to next	Active-passive disaster recovery
Weighted	Distributes by weight (1-1000); weight 300 gets 3x traffic vs weight 100	Canary deployments, gradual migration
Performance	Returns endpoint with lowest latency from client DNS resolver	Multi-region apps needing closest region
Geographic	Routes based on geographic origin of the DNS query	Data sovereignty and compliance requirements
Multivalue	Returns multiple healthy IPs in one DNS response; client picks	Simple client-side redundancy (external endpoints only)
Subnet	Maps specific client IP ranges to specific endpoints	Per-customer routing, VIP access, testing

Character Scenarios

☁️ Elena — Performance routing: Skyline Logistics’ customer portal is deployed in Australia East and West Europe. Traffic Manager with Performance routing sends Australian customers to the AU East endpoint and European customers to West Europe — lowest latency for each.

🏪 Sam — Priority routing: Harbour Retail has a primary web app in Australia East and a standby in Southeast Asia. Priority routing sends all traffic to AU East (priority 1). If it fails health checks, traffic automatically shifts to Southeast Asia (priority 2).

Endpoint Types

Endpoint Type	What It Points To	Example
Azure endpoint	Azure resources (App Service, Public IP, Cloud Service)	Sam’s web app in AU East
External endpoint	Non-Azure resources (IP address or FQDN)	Partner’s API at api.partner.com
Nested endpoint	Another Traffic Manager profile	Elena’s geographic parent with performance children

ℹ️ Nested Profiles — Combining Routing Methods

Nested profiles let you combine routing methods. A common pattern:

Parent profile: Geographic routing

• Europe → Child profile (Performance routing between West Europe and North Europe)
• Asia-Pacific → Child profile (Performance routing between AU East and Southeast Asia)
• Americas → Child profile (Performance routing between East US and West US)

This ensures:

1. Users are routed to the correct geographic region (compliance/sovereignty)
2. Within that region, they reach the lowest-latency endpoint (performance)

Configuration:

• Create child profiles with Performance routing and their regional endpoints
• Create parent profile with Geographic routing
• Add child profiles as “Nested” endpoints in the parent
• Set minimum child endpoints for the nested endpoint to be considered healthy

Health Monitoring

Traffic Manager continuously probes endpoints to determine health:

Setting	Description	Default
Protocol	HTTP, HTTPS, or TCP	HTTP
Port	Port to probe	80
Path	URL path for HTTP/HTTPS probes	/
Probing interval	How often to check	30 seconds (10 seconds for fast interval)
Tolerated failures	Failures before marking unhealthy	3
Probe timeout	Time to wait for response	10 seconds (5 seconds for fast interval)

Expected probe response: HTTP 200 OK. Any other response (or timeout) counts as a failure.

Fast interval probing: Reduces detection time from ~90 seconds (3 failures at 30s) to ~15 seconds (3 failures at 5s). Costs more but provides faster failover.

Custom health checks: For HTTPS probes, Traffic Manager validates the certificate chain. For HTTP probes, the response body can be checked with an expected status code range.

Key Takeaways

• Traffic Manager is DNS-based — it returns IPs, doesn’t proxy traffic
• Six routing methods cover DR (Priority), migration (Weighted), performance, geo, multivalue, and subnet
• Nested profiles combine routing methods for complex scenarios
• Health monitoring checks every 30 seconds by default; fast interval does 10 seconds
• DNS TTL affects failover time — lower TTL means faster client updates

---

Test Your Knowledge

Question

How does Traffic Manager route traffic?

Click or press Enter to reveal answer

Answer

At the DNS level. When a client queries the Traffic Manager FQDN, it returns the IP of the best endpoint based on the routing method. Traffic Manager never proxies or inspects the actual data traffic.

Click to flip back

Question

What's the difference between Performance and Geographic routing?

Click or press Enter to reveal answer

Answer

Performance: routes to the endpoint with lowest latency from the client. Geographic: routes based on the client's geographic location to a fixed endpoint. Performance optimises speed; Geographic enforces compliance/sovereignty.

Click to flip back

Question

What is a nested Traffic Manager profile?

Click or press Enter to reveal answer

Answer

A child Traffic Manager profile used as an endpoint in a parent profile. Allows combining routing methods — e.g., Geographic parent routing to Performance children for latency optimisation within each region.

Click to flip back

---

Knowledge Check

Elena needs users in Europe to always reach European endpoints (data sovereignty), but within Europe, traffic should go to the fastest endpoint. Which routing pattern achieves this?

APerformance routing only — it handles both requirementsBGeographic parent profile with nested Performance child profiles per regionCWeighted routing with higher weights for European endpointsDPriority routing with European endpoints as highest priority

Check Answer

Knowledge Check

Sam's Traffic Manager health probe checks every 30 seconds with 3 tolerated failures. How long before a failed endpoint is marked unhealthy?

A30 secondsB60 secondsCApproximately 90 secondsDApproximately 120 seconds

Check Answer

---

Next up: Application Gateway: Layer 7 — HTTP-aware load balancing with URL routing, TLS termination, and web application features.